1 # Ocean Interpreter - Stoney 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 second version of the interpreter exists to test out the
33 structured statement providing conditions and iteration, and simple
34 variable scoping. Clearly we need some minimal other functionality so
35 that values can be tested and instructions iterated over. All that
36 functionality is clearly not normative at this stage (not that
37 anything is **really** normative yet) and will change, so early test
38 code will certainly break in later versions.
40 The under-test parts of the language are:
42 - conditional/looping structured statements
43 - the `use` statement which is needed for that
44 - Variable binding using ":=" and "::=", and assignment using "=".
46 Elements which are present to make a usable language are:
48 - "blocks" of multiple statements.
49 - `pass`: a statement which does nothing.
50 - expressions: `+`, `-`, `*`, `/` can apply to numbers and `++` can
51 catenate strings. `and`, `or`, `not` manipulate Booleans, and
52 normal comparison operators can work on all three types.
53 - `print`: will print the values in a list of expressions.
54 - `program`: is given a list of identifiers to initialize from
59 Versions of the interpreter which obviously do not support a complete
60 language will be named after creeks and streams. This one is Stoney
63 Once we have something reasonably resembling a complete language, the
64 names of rivers will be used.
65 Early versions of the compiler will be named after seas. Major
66 releases of the compiler will be named after oceans. Hopefully I will
67 be finished once I get to the Pacific Ocean release.
71 As well as parsing and executing a program, the interpreter can print
72 out the program from the parsed internal structure. This is useful
73 for validating the parsing.
74 So the main requirements of the interpreter are:
76 - Parse the program, possibly with tracing,
77 - Analyse the parsed program to ensure consistency,
79 - Execute the program.
81 This is all performed by a single C program extracted with
84 There will be two formats for printing the program: a default and one
85 that uses bracketing. So a `--bracket` command line option is needed
86 for that. Normally the first code section found is used, however an
87 alternate section can be requested so that a file (such as this one)
88 can contain multiple programs This is effected with the `--section`
91 This code must be compiled with `-fplan9-extensions` so that anonymous
92 structures can be used.
94 ###### File: oceani.mk
96 myCFLAGS := -Wall -g -fplan9-extensions
97 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
98 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
99 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
101 all :: $(LDLIBS) oceani
102 oceani.c oceani.h : oceani.mdc parsergen
103 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
104 oceani.mk: oceani.mdc md2c
107 oceani: oceani.o $(LDLIBS)
108 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
110 ###### Parser: header
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, \
135 #include <sys/mman.h>
154 static char Usage[] = "Usage: oceani --trace --print --noexec --brackets"
155 "--section=SectionName prog.ocn\n";
156 static const struct option long_options[] = {
157 {"trace", 0, NULL, 't'},
158 {"print", 0, NULL, 'p'},
159 {"noexec", 0, NULL, 'n'},
160 {"brackets", 0, NULL, 'b'},
161 {"section", 1, NULL, 's'},
164 const char *options = "tpnbs";
165 int main(int argc, char *argv[])
171 char *section = NULL;
172 struct parse_context context = {
174 .ignored = (1 << TK_line_comment)
175 | (1 << TK_block_comment),
176 .number_chars = ".,_+-",
181 int doprint=0, dotrace=0, doexec=1, brackets=0;
184 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
187 case 't': dotrace=1; break;
188 case 'p': doprint=1; break;
189 case 'n': doexec=0; break;
190 case 'b': brackets=1; break;
191 case 's': section = optarg; break;
192 default: fprintf(stderr, Usage);
196 if (optind >= argc) {
197 fprintf(stderr, "oceani: no input file given\n");
200 fd = open(argv[optind], O_RDONLY);
202 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
205 context.file_name = argv[optind];
206 len = lseek(fd, 0, 2);
207 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
208 s = code_extract(file, file+len, NULL);
210 fprintf(stderr, "oceani: could not find any code in %s\n",
215 ## context initialization
219 for (ss = s; ss; ss = ss->next) {
220 struct text sec = ss->section;
221 if (sec.len == strlen(section) &&
222 strncmp(sec.txt, section, sec.len) == 0)
226 prog = parse_oceani(ss->code, &context.config,
227 dotrace ? stderr : NULL);
229 fprintf(stderr, "oceani: cannot find section %s\n",
234 prog = parse_oceani(s->code, &context.config,
235 dotrace ? stderr : NULL);
237 fprintf(stderr, "oceani: fatal parser error.\n");
238 context.parse_error = 1;
241 print_exec(*prog, 0, brackets);
242 if (prog && doexec && !context.parse_error) {
243 if (!analyse_prog(*prog, &context)) {
244 fprintf(stderr, "oceani: type error in program - not running.\n");
247 interp_prog(*prog, argv+optind+1);
254 struct section *t = s->next;
260 ## free context types
261 exit(context.parse_error ? 1 : 0);
266 The four requirements of parse, analyse, print, interpret apply to
267 each language element individually so that is how most of the code
270 Three of the four are fairly self explanatory. The one that requires
271 a little explanation is the analysis step.
273 The current language design does not require the types of variables to
274 be declared, but they must still have a single type. Different
275 operations impose different requirements on the variables, for example
276 addition requires both arguments to be numeric, and assignment
277 requires the variable on the left to have the same type as the
278 expression on the right.
280 Analysis involves propagating these type requirements around and
281 consequently setting the type of each variable. If any requirements
282 are violated (e.g. a string is compared with a number) or if a
283 variable needs to have two different types, then an error is raised
284 and the program will not run.
286 If the same variable is declared in both branchs of an 'if/else', or
287 in all cases of a 'switch' then the multiple instances may be merged
288 into just one variable if the variable is references after the
289 conditional statement. When this happens, the types must naturally be
290 consistent across all the branches. When the variable is not used
291 outside the if, the variables in the different branches are distinct
292 and can be of different types.
294 Determining the types of all variables early is important for
295 processing command line arguments. These can be assigned to any type
296 of variable, but we must first know the correct type so any required
297 conversion can happen. If a variable is associated with a command
298 line argument but no type can be interpreted (e.g. the variable is
299 only ever used in a `print` statement), then the type is set to
302 Undeclared names may only appear in "use" statements and "case" expressions.
303 These names are given a type of "label" and a unique value.
304 This allows them to fill the role of a name in an enumerated type, which
305 is useful for testing the `switch` statement.
307 As we will see, the condition part of a `while` statement can return
308 either a Boolean or some other type. This requires that the expect
309 type that gets passed around comprises a type (`enum vtype`) and a
310 flag to indicate that `Vbool` is also permitted.
312 As there are, as yet, no distinct types that are compatible, there
313 isn't much subtlety in the analysis. When we have distinct number
314 types, this will become more interesting.
318 When analysis discovers an inconsistency it needs to report an error;
319 just refusing to run the code ensures that the error doesn't cascade,
320 but by itself it isn't very useful. A clear understand of the sort of
321 error message that are useful will help guide the process of analysis.
323 At a simplistic level, the only sort of error that type analysis can
324 report is that the type of some construct doesn't match a contextual
325 requirement. For example, in `4 + "hello"` the addition provides a
326 contextual requirement for numbers, but `"hello"` is not a number. In
327 this particular example no further information is needed as the types
328 are obvious from local information. When a variable is involved that
329 isn't the case. It may be helpful to explain why the variable has a
330 particular type, by indicating the location where the type was set,
331 whether by declaration or usage.
333 Using a recursive-descent analysis we can easily detect a problem at
334 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
335 will detect that one argument is not a number and the usage of `hello`
336 will detect that a number was wanted, but not provided. In this
337 (early) version of the language, we will generate error reports at
338 multiple locations, so the use of `hello` will report an error and
339 explain were the value was set, and the addition will report an error
340 and say why numbers are needed. To be able to report locations for
341 errors, each language element will need to record a file location
342 (line and column) and each variable will need to record the language
343 element where its type was set. For now we will assume that each line
344 of an error message indicates one location in the file, and up to 2
345 types. So we provide a `printf`-like function which takes a format, a
346 language (a `struct exec` which has not yet been introduced), and 2
347 types. "`%1`" reports the first type, "`%2`" reports the second. We
348 will need a function to print the location, once we know how that is
349 stored. As will be explained later, there are sometimes extra rules for
350 type matching and they might affect error messages, we need to pass those
353 As well as type errors, we sometimes need to report problems with
354 tokens, which might be unexpected or might name a type that has not
355 been defined. For these we have `tok_err()` which reports an error
356 with a given token. Each of the error functions sets the flag in the
357 context so indicate that parsing failed.
361 static void fput_loc(struct exec *loc, FILE *f);
363 ###### core functions
365 static void type_err(struct parse_context *c,
366 char *fmt, struct exec *loc,
367 struct type *t1, int rules, struct type *t2)
369 fprintf(stderr, "%s:", c->file_name);
370 fput_loc(loc, stderr);
371 for (; *fmt ; fmt++) {
378 case '%': fputc(*fmt, stderr); break;
379 default: fputc('?', stderr); break;
382 fprintf(stderr, "%.*s", t1->name.len, t1->name.txt);
384 fputs("*unknown*", stderr);
388 fprintf(stderr, "%.*s", t2->name.len, t2->name.txt);
390 fputs("*unknown*", stderr);
400 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
402 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
403 t->txt.len, t->txt.txt);
409 One last introductory step before detailing the language elements and
410 providing their four requirements is to establish the data structures
411 to store these elements.
413 There are two key objects that we need to work with: executable
414 elements which comprise the program, and values which the program
415 works with. Between these are the variables in their various scopes
416 which hold the values, and types which classify the values stored and
417 manipulatd by executables.
421 Values come in a wide range of types, with more likely to be added.
422 Each type needs to be able to parse and print its own values (for
423 convenience at least) as well as to compare two values, at least for
424 equality and possibly for order. For now, values might need to be
425 duplicated and freed, though eventually such manipulations will be
426 better integrated into the language.
428 Rather than requiring every numeric type to support all numeric
429 operations (add, multiple, etc), we allow types to be able to present
430 as one of a few standard types: integer, float, and fraction. The
431 existance of these conversion functions enable types to determine if
432 they are compatible with other types.
434 Named type are stored in a simple linked list. Objects of each type are "values"
435 which are often passed around by value.
442 ## value union fields
449 struct value (*init)(struct type *type);
450 struct value (*prepare)(struct type *type);
451 struct value (*parse)(struct type *type, char *str);
452 void (*print)(struct value val);
453 int (*cmp_order)(struct value v1, struct value v2);
454 int (*cmp_eq)(struct value v1, struct value v2);
455 struct value (*dup)(struct value val);
456 void (*free)(struct value val);
457 struct type *(*compat)(struct type *this, struct type *other);
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);
468 struct type *typelist;
472 static struct type *find_type(struct parse_context *c, struct text s)
474 struct type *l = c->typelist;
477 text_cmp(l->name, s) != 0)
482 static struct type *add_type(struct parse_context *c, struct text s,
487 n = calloc(1, sizeof(*n));
490 n->next = c->typelist;
495 static void free_type(struct type *t)
497 /* The type is always a reference to something in the
498 * context, so we don't need to free anything.
502 static void free_value(struct value v)
508 static struct value val_prepare(struct type *type)
513 return type->prepare(type);
518 static struct value val_init(struct type *type)
523 return type->init(type);
528 static struct value dup_value(struct value v)
531 return v.type->dup(v);
535 static int value_cmp(struct value left, struct value right)
537 if (left.type && left.type->cmp_order)
538 return left.type->cmp_order(left, right);
539 if (left.type && left.type->cmp_eq)
540 return left.type->cmp_eq(left, right);
544 static void print_value(struct value v)
546 if (v.type && v.type->print)
552 static struct value parse_value(struct type *type, char *arg)
556 if (type && type->parse)
557 return type->parse(type, arg);
562 ###### free context types
564 while (context.typelist) {
565 struct type *t = context.typelist;
567 context.typelist = t->next;
573 Values of the base types can be numbers, which we represent as
574 multi-precision fractions, strings, Booleans and labels. When
575 analysing the program we also need to allow for places where no value
576 is meaningful (type `Tnone`) and where we don't know what type to
577 expect yet (type is `NULL`).
579 Values are never shared, they are always copied when used, and freed
580 when no longer needed.
582 When propagating type information around the program, we need to
583 determine if two types are compatible, where type `NULL` is compatible
584 with anything. There are two special cases with type compatibility,
585 both related to the Conditional Statement which will be described
586 later. In some cases a Boolean can be accepted as well as some other
587 primary type, and in others any type is acceptable except a label (`Vlabel`).
588 A separate function encode these cases will simplify some code later.
590 When assigning command line arguments to variables, we need to be able
591 to parse each type from a string.
599 myLDLIBS := libnumber.o libstring.o -lgmp
600 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
602 ###### type union fields
603 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
605 ###### value union fields
612 static void _free_value(struct value v)
614 switch (v.type->vtype) {
616 case Vstr: free(v.str.txt); break;
617 case Vnum: mpq_clear(v.num); break;
623 static int vtype_compat(struct type *require, struct type *have, int rules)
625 if ((rules & Rboolok) && have == Tbool)
627 if ((rules & Rnolabel) && have == Tlabel)
629 if (!require || !have)
632 return require == have;
635 ###### value functions
637 static struct value _val_prepare(struct type *type)
642 switch(type->vtype) {
646 memset(&rv.num, 0, sizeof(rv.num));
662 static struct value _val_init(struct type *type)
667 switch(type->vtype) {
671 mpq_init(rv.num); break;
673 rv.str.txt = malloc(1);
686 static struct value _dup_value(struct value v)
690 switch (rv.type->vtype) {
701 mpq_set(rv.num, v.num);
704 rv.str.len = v.str.len;
705 rv.str.txt = malloc(rv.str.len);
706 memcpy(rv.str.txt, v.str.txt, v.str.len);
712 static int _value_cmp(struct value left, struct value right)
715 if (left.type != right.type)
716 return left.type - right.type;
717 switch (left.type->vtype) {
718 case Vlabel: cmp = left.label == right.label ? 0 : 1; break;
719 case Vnum: cmp = mpq_cmp(left.num, right.num); break;
720 case Vstr: cmp = text_cmp(left.str, right.str); break;
721 case Vbool: cmp = left.bool - right.bool; break;
727 static void _print_value(struct value v)
729 switch (v.type->vtype) {
731 printf("*no-value*"); break;
733 printf("*label-%p*", v.label); break;
735 printf("%.*s", v.str.len, v.str.txt); break;
737 printf("%s", v.bool ? "True":"False"); break;
742 mpf_set_q(fl, v.num);
743 gmp_printf("%Fg", fl);
750 static struct value _parse_value(struct type *type, char *arg)
758 switch(type->vtype) {
764 val.str.len = strlen(arg);
765 val.str.txt = malloc(val.str.len);
766 memcpy(val.str.txt, arg, val.str.len);
773 tx.txt = arg; tx.len = strlen(tx.txt);
774 if (number_parse(val.num, tail, tx) == 0)
777 mpq_neg(val.num, val.num);
779 printf("Unsupported suffix: %s\n", arg);
784 if (strcasecmp(arg, "true") == 0 ||
785 strcmp(arg, "1") == 0)
787 else if (strcasecmp(arg, "false") == 0 ||
788 strcmp(arg, "0") == 0)
791 printf("Bad bool: %s\n", arg);
799 static void _free_value(struct value v);
801 static struct type base_prototype = {
803 .prepare = _val_prepare,
804 .parse = _parse_value,
805 .print = _print_value,
806 .cmp_order = _value_cmp,
807 .cmp_eq = _value_cmp,
812 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
815 static struct type *add_base_type(struct parse_context *c, char *n, enum vtype vt)
817 struct text txt = { n, strlen(n) };
820 t = add_type(c, txt, &base_prototype);
825 ###### context initialization
827 Tbool = add_base_type(&context, "Boolean", Vbool);
828 Tstr = add_base_type(&context, "string", Vstr);
829 Tnum = add_base_type(&context, "number", Vnum);
830 Tnone = add_base_type(&context, "none", Vnone);
831 Tlabel = add_base_type(&context, "label", Vlabel);
835 Variables are scoped named values. We store the names in a linked
836 list of "bindings" sorted lexically, and use sequential search and
843 struct binding *next; // in lexical order
847 This linked list is stored in the parse context so that "reduce"
848 functions can find or add variables, and so the analysis phase can
849 ensure that every variable gets a type.
853 struct binding *varlist; // In lexical order
857 static struct binding *find_binding(struct parse_context *c, struct text s)
859 struct binding **l = &c->varlist;
864 (cmp = text_cmp((*l)->name, s)) < 0)
868 n = calloc(1, sizeof(*n));
875 Each name can be linked to multiple variables defined in different
876 scopes. Each scope starts where the name is declared and continues
877 until the end of the containing code block. Scopes of a given name
878 cannot nest, so a declaration while a name is in-scope is an error.
880 ###### binding fields
881 struct variable *var;
885 struct variable *previous;
887 struct binding *name;
888 struct exec *where_decl;// where name was declared
889 struct exec *where_set; // where type was set
893 While the naming seems strange, we include local constants in the
894 definition of variables. A name declared `var := value` can
895 subsequently be changed, but a name declared `var ::= value` cannot -
898 ###### variable fields
901 Scopes in parallel branches can be partially merged. More
902 specifically, if a given name is declared in both branches of an
903 if/else then its scope is a candidate for merging. Similarly if
904 every branch of an exhaustive switch (e.g. has an "else" clause)
905 declares a given name, then the scopes from the branches are
906 candidates for merging.
908 Note that names declared inside a loop (which is only parallel to
909 itself) are never visible after the loop. Similarly names defined in
910 scopes which are not parallel, such as those started by `for` and
911 `switch`, are never visible after the scope. Only variables defined in
912 both `then` and `else` (including the implicit then after an `if`, and
913 excluding `then` used with `for`) and in all `case`s and `else` of a
914 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
916 Labels, which are a bit like variables, follow different rules.
917 Labels are not explicitly declared, but if an undeclared name appears
918 in a context where a label is legal, that effectively declares the
919 name as a label. The declaration remains in force (or in scope) at
920 least to the end of the immediately containing block and conditionally
921 in any larger containing block which does not declare the name in some
922 other way. Importantly, the conditional scope extension happens even
923 if the label is only used in one parallel branch of a conditional --
924 when used in one branch it is treated as having been declared in all
927 Merge candidates are tentatively visible beyond the end of the
928 branching statement which creates them. If the name is used, the
929 merge is affirmed and they become a single variable visible at the
930 outer layer. If not - if it is redeclared first - the merge lapses.
932 To track scopes we have an extra stack, implemented as a linked list,
933 which roughly parallels the parse stack and which is used exclusively
934 for scoping. When a new scope is opened, a new frame is pushed and
935 the child-count of the parent frame is incremented. This child-count
936 is used to distinguish between the first of a set of parallel scopes,
937 in which declared variables must not be in scope, and subsequent
938 branches, whether they must already be conditionally scoped.
940 To push a new frame *before* any code in the frame is parsed, we need a
941 grammar reduction. This is most easily achieved with a grammar
942 element which derives the empty string, and creates the new scope when
943 it is recognized. This can be placed, for example, between a keyword
944 like "if" and the code following it.
948 struct scope *parent;
954 struct scope *scope_stack;
957 static void scope_pop(struct parse_context *c)
959 struct scope *s = c->scope_stack;
961 c->scope_stack = s->parent;
966 static void scope_push(struct parse_context *c)
968 struct scope *s = calloc(1, sizeof(*s));
970 c->scope_stack->child_count += 1;
971 s->parent = c->scope_stack;
979 OpenScope -> ${ scope_push(config2context(config)); }$
982 Each variable records a scope depth and is in one of four states:
984 - "in scope". This is the case between the declaration of the
985 variable and the end of the containing block, and also between
986 the usage with affirms a merge and the end of that block.
988 The scope depth is not greater than the current parse context scope
989 nest depth. When the block of that depth closes, the state will
990 change. To achieve this, all "in scope" variables are linked
991 together as a stack in nesting order.
993 - "pending". The "in scope" block has closed, but other parallel
994 scopes are still being processed. So far, every parallel block at
995 the same level that has closed has declared the name.
997 The scope depth is the depth of the last parallel block that
998 enclosed the declaration, and that has closed.
1000 - "conditionally in scope". The "in scope" block and all parallel
1001 scopes have closed, and no further mention of the name has been
1002 seen. This state includes a secondary nest depth which records the
1003 outermost scope seen since the variable became conditionally in
1004 scope. If a use of the name is found, the variable becomes "in
1005 scope" and that secondary depth becomes the recorded scope depth.
1006 If the name is declared as a new variable, the old variable becomes
1007 "out of scope" and the recorded scope depth stays unchanged.
1009 - "out of scope". The variable is neither in scope nor conditionally
1010 in scope. It is permanently out of scope now and can be removed from
1011 the "in scope" stack.
1014 ###### variable fields
1015 int depth, min_depth;
1016 enum { OutScope, PendingScope, CondScope, InScope } scope;
1017 struct variable *in_scope;
1019 ###### parse context
1021 struct variable *in_scope;
1023 All variables with the same name are linked together using the
1024 'previous' link. Those variable that have
1025 been affirmatively merged all have a 'merged' pointer that points to
1026 one primary variable - the most recently declared instance. When
1027 merging variables, we need to also adjust the 'merged' pointer on any
1028 other variables that had previously been merged with the one that will
1029 no longer be primary.
1031 ###### variable fields
1032 struct variable *merged;
1034 ###### ast functions
1036 static void variable_merge(struct variable *primary, struct variable *secondary)
1040 if (primary->merged)
1042 primary = primary->merged;
1044 for (v = primary->previous; v; v=v->previous)
1045 if (v == secondary || v == secondary->merged ||
1046 v->merged == secondary ||
1047 (v->merged && v->merged == secondary->merged)) {
1048 v->scope = OutScope;
1049 v->merged = primary;
1053 ###### free context vars
1055 while (context.varlist) {
1056 struct binding *b = context.varlist;
1057 struct variable *v = b->var;
1058 context.varlist = b->next;
1061 struct variable *t = v;
1069 #### Manipulating Bindings
1071 When a name is conditionally visible, a new declaration discards the
1072 old binding - the condition lapses. Conversely a usage of the name
1073 affirms the visibility and extends it to the end of the containing
1074 block - i.e. the block that contains both the original declaration and
1075 the latest usage. This is determined from `min_depth`. When a
1076 conditionally visible variable gets affirmed like this, it is also
1077 merged with other conditionally visible variables with the same name.
1079 When we parse a variable declaration we either signal an error if the
1080 name is currently bound, or create a new variable at the current nest
1081 depth if the name is unbound or bound to a conditionally scoped or
1082 pending-scope variable. If the previous variable was conditionally
1083 scoped, it and its homonyms becomes out-of-scope.
1085 When we parse a variable reference (including non-declarative
1086 assignment) we signal an error if the name is not bound or is bound to
1087 a pending-scope variable; update the scope if the name is bound to a
1088 conditionally scoped variable; or just proceed normally if the named
1089 variable is in scope.
1091 When we exit a scope, any variables bound at this level are either
1092 marked out of scope or pending-scoped, depending on whether the
1093 scope was sequential or parallel.
1095 When exiting a parallel scope we check if there are any variables that
1096 were previously pending and are still visible. If there are, then
1097 there weren't redeclared in the most recent scope, so they cannot be
1098 merged and must become out-of-scope. If it is not the first of
1099 parallel scopes (based on `child_count`), we check that there was a
1100 previous binding that is still pending-scope. If there isn't, the new
1101 variable must now be out-of-scope.
1103 When exiting a sequential scope that immediately enclosed parallel
1104 scopes, we need to resolve any pending-scope variables. If there was
1105 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1106 we need to mark all pending-scope variable as out-of-scope. Otherwise
1107 all pending-scope variables become conditionally scoped.
1110 enum closetype { CloseSequential, CloseParallel, CloseElse };
1112 ###### ast functions
1114 static struct variable *var_decl(struct parse_context *c, struct text s)
1116 struct binding *b = find_binding(c, s);
1117 struct variable *v = b->var;
1119 switch (v ? v->scope : OutScope) {
1121 /* Caller will report the error */
1125 v && v->scope == CondScope;
1127 v->scope = OutScope;
1131 v = calloc(1, sizeof(*v));
1132 v->previous = b->var;
1135 v->min_depth = v->depth = c->scope_depth;
1137 v->in_scope = c->in_scope;
1139 v->val = val_prepare(NULL);
1143 static struct variable *var_ref(struct parse_context *c, struct text s)
1145 struct binding *b = find_binding(c, s);
1146 struct variable *v = b->var;
1147 struct variable *v2;
1149 switch (v ? v->scope : OutScope) {
1152 /* Signal an error - once that is possible */
1155 /* All CondScope variables of this name need to be merged
1156 * and become InScope
1158 v->depth = v->min_depth;
1160 for (v2 = v->previous;
1161 v2 && v2->scope == CondScope;
1163 variable_merge(v, v2);
1171 static void var_block_close(struct parse_context *c, enum closetype ct)
1173 /* close of all variables that are in_scope */
1174 struct variable *v, **vp, *v2;
1177 for (vp = &c->in_scope;
1178 v = *vp, v && v->depth > c->scope_depth && v->min_depth > c->scope_depth;
1182 case CloseParallel: /* handle PendingScope */
1186 if (c->scope_stack->child_count == 1)
1187 v->scope = PendingScope;
1188 else if (v->previous &&
1189 v->previous->scope == PendingScope)
1190 v->scope = PendingScope;
1191 else if (v->val.type == Tlabel)
1192 v->scope = PendingScope;
1193 else if (v->name->var == v)
1194 v->scope = OutScope;
1195 if (ct == CloseElse) {
1196 /* All Pending variables with this name
1197 * are now Conditional */
1199 v2 && v2->scope == PendingScope;
1201 v2->scope = CondScope;
1206 v2 && v2->scope == PendingScope;
1208 if (v2->val.type != Tlabel)
1209 v2->scope = OutScope;
1211 case OutScope: break;
1214 case CloseSequential:
1215 if (v->val.type == Tlabel)
1216 v->scope = PendingScope;
1219 v->scope = OutScope;
1222 /* There was no 'else', so we can only become
1223 * conditional if we know the cases were exhaustive,
1224 * and that doesn't mean anything yet.
1225 * So only labels become conditional..
1228 v2 && v2->scope == PendingScope;
1230 if (v2->val.type == Tlabel) {
1231 v2->scope = CondScope;
1232 v2->min_depth = c->scope_depth;
1234 v2->scope = OutScope;
1237 case OutScope: break;
1241 if (v->scope == OutScope)
1250 Executables can be lots of different things. In many cases an
1251 executable is just an operation combined with one or two other
1252 executables. This allows for expressions and lists etc. Other times
1253 an executable is something quite specific like a constant or variable
1254 name. So we define a `struct exec` to be a general executable with a
1255 type, and a `struct binode` which is a subclass of `exec`, forms a
1256 node in a binary tree, and holds an operation. There will be other
1257 subclasses, and to access these we need to be able to `cast` the
1258 `exec` into the various other types.
1261 #define cast(structname, pointer) ({ \
1262 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
1263 if (__mptr && *__mptr != X##structname) abort(); \
1264 (struct structname *)( (char *)__mptr);})
1266 #define new(structname) ({ \
1267 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1268 __ptr->type = X##structname; \
1269 __ptr->line = -1; __ptr->column = -1; \
1272 #define new_pos(structname, token) ({ \
1273 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1274 __ptr->type = X##structname; \
1275 __ptr->line = token.line; __ptr->column = token.col; \
1284 enum exec_types type;
1292 struct exec *left, *right;
1295 ###### ast functions
1297 static int __fput_loc(struct exec *loc, FILE *f)
1299 if (loc->line >= 0) {
1300 fprintf(f, "%d:%d: ", loc->line, loc->column);
1303 if (loc->type == Xbinode)
1304 return __fput_loc(cast(binode,loc)->left, f) ||
1305 __fput_loc(cast(binode,loc)->right, f);
1308 static void fput_loc(struct exec *loc, FILE *f)
1310 if (!__fput_loc(loc, f))
1311 fprintf(f, "??:??: ");
1314 Each different type of `exec` node needs a number of functions
1315 defined, a bit like methods. We must be able to be able to free it,
1316 print it, analyse it and execute it. Once we have specific `exec`
1317 types we will need to parse them too. Let's take this a bit more
1322 The parser generator requires a `free_foo` function for each struct
1323 that stores attributes and they will be `exec`s and subtypes there-of.
1324 So we need `free_exec` which can handle all the subtypes, and we need
1327 ###### ast functions
1329 static void free_binode(struct binode *b)
1334 free_exec(b->right);
1338 ###### core functions
1339 static void free_exec(struct exec *e)
1348 ###### forward decls
1350 static void free_exec(struct exec *e);
1352 ###### free exec cases
1353 case Xbinode: free_binode(cast(binode, e)); break;
1357 Printing an `exec` requires that we know the current indent level for
1358 printing line-oriented components. As will become clear later, we
1359 also want to know what sort of bracketing to use.
1361 ###### ast functions
1363 static void do_indent(int i, char *str)
1370 ###### core functions
1371 static void print_binode(struct binode *b, int indent, int bracket)
1375 ## print binode cases
1379 static void print_exec(struct exec *e, int indent, int bracket)
1385 print_binode(cast(binode, e), indent, bracket); break;
1390 ###### forward decls
1392 static void print_exec(struct exec *e, int indent, int bracket);
1396 As discussed, analysis involves propagating type requirements around
1397 the program and looking for errors.
1399 So `propagate_types` is passed an expected type (being a `struct type`
1400 pointer together with some `val_rules` flags) that the `exec` is
1401 expected to return, and returns the type that it does return, either
1402 of which can be `NULL` signifying "unknown". An `ok` flag is passed
1403 by reference. It is set to `0` when an error is found, and `2` when
1404 any change is made. If it remains unchanged at `1`, then no more
1405 propagation is needed.
1409 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1};
1413 if (rules & Rnolabel)
1414 fputs(" (labels not permitted)", stderr);
1417 ###### core functions
1419 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1420 struct type *type, int rules)
1427 switch (prog->type) {
1430 struct binode *b = cast(binode, prog);
1432 ## propagate binode cases
1436 ## propagate exec cases
1443 Interpreting an `exec` doesn't require anything but the `exec`. State
1444 is stored in variables and each variable will be directly linked from
1445 within the `exec` tree. The exception to this is the whole `program`
1446 which needs to look at command line arguments. The `program` will be
1447 interpreted separately.
1449 Each `exec` can return a value, which may be `Tnone` but must be non-NULL;
1451 ###### core functions
1453 static struct value interp_exec(struct exec *e)
1463 struct binode *b = cast(binode, e);
1464 struct value left, right;
1465 left.type = right.type = Tnone;
1467 ## interp binode cases
1469 free_value(left); free_value(right);
1472 ## interp exec cases
1477 ## Language elements
1479 Each language element needs to be parsed, printed, analysed,
1480 interpreted, and freed. There are several, so let's just start with
1481 the easy ones and work our way up.
1485 We have already met values as separate objects. When manifest
1486 constants appear in the program text, that must result in an executable
1487 which has a constant value. So the `val` structure embeds a value in
1503 $0 = new_pos(val, $1);
1504 $0->val.type = Tbool;
1508 $0 = new_pos(val, $1);
1509 $0->val.type = Tbool;
1513 $0 = new_pos(val, $1);
1514 $0->val.type = Tnum;
1517 if (number_parse($0->val.num, tail, $1.txt) == 0)
1518 mpq_init($0->val.num);
1520 tok_err(config2context(config), "error: unsupported number suffix",
1525 $0 = new_pos(val, $1);
1526 $0->val.type = Tstr;
1529 string_parse(&$1, '\\', &$0->val.str, tail);
1531 tok_err(config2context(config), "error: unsupported string suffix",
1536 $0 = new_pos(val, $1);
1537 $0->val.type = Tstr;
1540 string_parse(&$1, '\\', &$0->val.str, tail);
1542 tok_err(config2context(config), "error: unsupported string suffix",
1547 ###### print exec cases
1550 struct val *v = cast(val, e);
1551 if (v->val.type == Tstr)
1553 print_value(v->val);
1554 if (v->val.type == Tstr)
1559 ###### propagate exec cases
1562 struct val *val = cast(val, prog);
1563 if (!vtype_compat(type, val->val.type, rules)) {
1564 type_err(c, "error: expected %1%r found %2",
1565 prog, type, rules, val->val.type);
1568 return val->val.type;
1571 ###### interp exec cases
1573 return dup_value(cast(val, e)->val);
1575 ###### ast functions
1576 static void free_val(struct val *v)
1584 ###### free exec cases
1585 case Xval: free_val(cast(val, e)); break;
1587 ###### ast functions
1588 // Move all nodes from 'b' to 'rv', reversing the order.
1589 // In 'b' 'left' is a list, and 'right' is the last node.
1590 // In 'rv', left' is the first node and 'right' is a list.
1591 static struct binode *reorder_bilist(struct binode *b)
1593 struct binode *rv = NULL;
1596 struct exec *t = b->right;
1600 b = cast(binode, b->left);
1610 Just as we used a `val` to wrap a value into an `exec`, we similarly
1611 need a `var` to wrap a `variable` into an exec. While each `val`
1612 contained a copy of the value, each `var` hold a link to the variable
1613 because it really is the same variable no matter where it appears.
1614 When a variable is used, we need to remember to follow the `->merged`
1615 link to find the primary instance.
1623 struct variable *var;
1629 VariableDecl -> IDENTIFIER : ${ {
1630 struct variable *v = var_decl(config2context(config), $1.txt);
1631 $0 = new_pos(var, $1);
1636 v = var_ref(config2context(config), $1.txt);
1638 type_err(config2context(config), "error: variable '%v' redeclared",
1639 $0, Tnone, 0, Tnone);
1640 type_err(config2context(config), "info: this is where '%v' was first declared",
1641 v->where_decl, Tnone, 0, Tnone);
1644 | IDENTIFIER :: ${ {
1645 struct variable *v = var_decl(config2context(config), $1.txt);
1646 $0 = new_pos(var, $1);
1652 v = var_ref(config2context(config), $1.txt);
1654 type_err(config2context(config), "error: variable '%v' redeclared",
1655 $0, Tnone, 0, Tnone);
1656 type_err(config2context(config), "info: this is where '%v' was first declared",
1657 v->where_decl, Tnone, 0, Tnone);
1660 | IDENTIFIER : Type ${ {
1661 struct variable *v = var_decl(config2context(config), $1.txt);
1662 $0 = new_pos(var, $1);
1667 v->val = val_prepare($<3);
1669 v = var_ref(config2context(config), $1.txt);
1671 type_err(config2context(config), "error: variable '%v' redeclared",
1672 $0, Tnone, 0, Tnone);
1673 type_err(config2context(config), "info: this is where '%v' was first declared",
1674 v->where_decl, Tnone, 0, Tnone);
1677 | IDENTIFIER :: Type ${ {
1678 struct variable *v = var_decl(config2context(config), $1.txt);
1679 $0 = new_pos(var, $1);
1684 v->val = val_prepare($<3);
1687 v = var_ref(config2context(config), $1.txt);
1689 type_err(config2context(config), "error: variable '%v' redeclared",
1690 $0, Tnone, 0, Tnone);
1691 type_err(config2context(config), "info: this is where '%v' was first declared",
1692 v->where_decl, Tnone, 0, Tnone);
1696 Variable -> IDENTIFIER ${ {
1697 struct variable *v = var_ref(config2context(config), $1.txt);
1698 $0 = new_pos(var, $1);
1700 /* This might be a label - allocate a var just in case */
1701 v = var_decl(config2context(config), $1.txt);
1703 v->val = val_prepare(Tlabel);
1704 v->val.label = &v->val;
1712 Type -> IDENTIFIER ${
1713 $0 = find_type(config2context(config), $1.txt);
1715 tok_err(config2context(config),
1716 "error: undefined type", &$1);
1722 ###### print exec cases
1725 struct var *v = cast(var, e);
1727 struct binding *b = v->var->name;
1728 printf("%.*s", b->name.len, b->name.txt);
1735 if (loc->type == Xvar) {
1736 struct var *v = cast(var, loc);
1738 struct binding *b = v->var->name;
1739 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
1741 fputs("???", stderr);
1743 fputs("NOTVAR", stderr);
1746 ###### propagate exec cases
1750 struct var *var = cast(var, prog);
1751 struct variable *v = var->var;
1753 type_err(c, "%d:BUG: no variable!!", prog, Tnone, 0, Tnone);
1759 if (v->val.type == NULL) {
1760 if (type && *ok != 0) {
1761 v->val = val_prepare(type);
1762 v->where_set = prog;
1767 if (!vtype_compat(type, v->val.type, rules)) {
1768 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
1769 type, rules, v->val.type);
1770 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
1771 v->val.type, rules, Tnone);
1779 ###### interp exec cases
1782 struct var *var = cast(var, e);
1783 struct variable *v = var->var;
1787 return dup_value(v->val);
1790 ###### ast functions
1792 static void free_var(struct var *v)
1797 ###### free exec cases
1798 case Xvar: free_var(cast(var, e)); break;
1800 ### Expressions: Boolean
1802 Our first user of the `binode` will be expressions, and particularly
1803 Boolean expressions. As I haven't implemented precedence in the
1804 parser generator yet, we need different names for each precedence
1805 level used by expressions. The outer most or lowest level precedence
1806 are Boolean `or` `and`, and `not` which form an `Expression` out of `BTerm`s
1817 Expression -> Expression or BTerm ${ {
1818 struct binode *b = new(binode);
1824 | BTerm ${ $0 = $<1; }$
1826 BTerm -> BTerm and BFact ${ {
1827 struct binode *b = new(binode);
1833 | BFact ${ $0 = $<1; }$
1835 BFact -> not BFact ${ {
1836 struct binode *b = new(binode);
1843 ###### print binode cases
1845 print_exec(b->left, -1, 0);
1847 print_exec(b->right, -1, 0);
1850 print_exec(b->left, -1, 0);
1852 print_exec(b->right, -1, 0);
1856 print_exec(b->right, -1, 0);
1859 ###### propagate binode cases
1863 /* both must be Tbool, result is Tbool */
1864 propagate_types(b->left, c, ok, Tbool, 0);
1865 propagate_types(b->right, c, ok, Tbool, 0);
1866 if (type && type != Tbool) {
1867 type_err(c, "error: %1 operation found where %2 expected", prog,
1873 ###### interp binode cases
1875 rv = interp_exec(b->left);
1876 right = interp_exec(b->right);
1877 rv.bool = rv.bool && right.bool;
1880 rv = interp_exec(b->left);
1881 right = interp_exec(b->right);
1882 rv.bool = rv.bool || right.bool;
1885 rv = interp_exec(b->right);
1889 ### Expressions: Comparison
1891 Of slightly higher precedence that Boolean expressions are
1893 A comparison takes arguments of any type, but the two types must be
1896 To simplify the parsing we introduce an `eop` which can record an
1897 expression operator.
1904 ###### ast functions
1905 static void free_eop(struct eop *e)
1920 | Expr CMPop Expr ${ {
1921 struct binode *b = new(binode);
1927 | Expr ${ $0 = $<1; }$
1932 CMPop -> < ${ $0.op = Less; }$
1933 | > ${ $0.op = Gtr; }$
1934 | <= ${ $0.op = LessEq; }$
1935 | >= ${ $0.op = GtrEq; }$
1936 | == ${ $0.op = Eql; }$
1937 | != ${ $0.op = NEql; }$
1939 ###### print binode cases
1947 print_exec(b->left, -1, 0);
1949 case Less: printf(" < "); break;
1950 case LessEq: printf(" <= "); break;
1951 case Gtr: printf(" > "); break;
1952 case GtrEq: printf(" >= "); break;
1953 case Eql: printf(" == "); break;
1954 case NEql: printf(" != "); break;
1957 print_exec(b->right, -1, 0);
1960 ###### propagate binode cases
1967 /* Both must match but not be labels, result is Tbool */
1968 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
1970 propagate_types(b->right, c, ok, t, 0);
1972 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
1974 t = propagate_types(b->left, c, ok, t, 0);
1976 if (!vtype_compat(type, Tbool, 0)) {
1977 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
1978 Tbool, rules, type);
1983 ###### interp binode cases
1992 left = interp_exec(b->left);
1993 right = interp_exec(b->right);
1994 cmp = value_cmp(left, right);
1997 case Less: rv.bool = cmp < 0; break;
1998 case LessEq: rv.bool = cmp <= 0; break;
1999 case Gtr: rv.bool = cmp > 0; break;
2000 case GtrEq: rv.bool = cmp >= 0; break;
2001 case Eql: rv.bool = cmp == 0; break;
2002 case NEql: rv.bool = cmp != 0; break;
2003 default: rv.bool = 0; break;
2008 ### Expressions: The rest
2010 The remaining expressions with the highest precedence are arithmetic
2011 and string concatenation. They are `Expr`, `Term`, and `Factor`.
2012 The `Factor` is where the `Value` and `Variable` that we already have
2015 `+` and `-` are both infix and prefix operations (where they are
2016 absolute value and negation). These have different operator names.
2018 We also have a 'Bracket' operator which records where parentheses were
2019 found. This makes it easy to reproduce these when printing. Once
2020 precedence is handled better I might be able to discard this.
2032 Expr -> Expr Eop Term ${ {
2033 struct binode *b = new(binode);
2039 | Term ${ $0 = $<1; }$
2041 Term -> Term Top Factor ${ {
2042 struct binode *b = new(binode);
2048 | Factor ${ $0 = $<1; }$
2050 Factor -> ( Expression ) ${ {
2051 struct binode *b = new_pos(binode, $1);
2057 struct binode *b = new(binode);
2062 | Value ${ $0 = $<1; }$
2063 | Variable ${ $0 = $<1; }$
2066 Eop -> + ${ $0.op = Plus; }$
2067 | - ${ $0.op = Minus; }$
2069 Uop -> + ${ $0.op = Absolute; }$
2070 | - ${ $0.op = Negate; }$
2072 Top -> * ${ $0.op = Times; }$
2073 | / ${ $0.op = Divide; }$
2074 | ++ ${ $0.op = Concat; }$
2076 ###### print binode cases
2082 print_exec(b->left, indent, 0);
2084 case Plus: printf(" + "); break;
2085 case Minus: printf(" - "); break;
2086 case Times: printf(" * "); break;
2087 case Divide: printf(" / "); break;
2088 case Concat: printf(" ++ "); break;
2091 print_exec(b->right, indent, 0);
2095 print_exec(b->right, indent, 0);
2099 print_exec(b->right, indent, 0);
2103 print_exec(b->right, indent, 0);
2107 ###### propagate binode cases
2112 /* both must be numbers, result is Tnum */
2115 /* as propagate_types ignores a NULL,
2116 * unary ops fit here too */
2117 propagate_types(b->left, c, ok, Tnum, 0);
2118 propagate_types(b->right, c, ok, Tnum, 0);
2119 if (!vtype_compat(type, Tnum, 0)) {
2120 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
2127 /* both must be Tstr, result is Tstr */
2128 propagate_types(b->left, c, ok, Tstr, 0);
2129 propagate_types(b->right, c, ok, Tstr, 0);
2130 if (!vtype_compat(type, Tstr, 0)) {
2131 type_err(c, "error: Concat returns %1 but %2 expected", prog,
2138 return propagate_types(b->right, c, ok, type, 0);
2140 ###### interp binode cases
2143 rv = interp_exec(b->left);
2144 right = interp_exec(b->right);
2145 mpq_add(rv.num, rv.num, right.num);
2148 rv = interp_exec(b->left);
2149 right = interp_exec(b->right);
2150 mpq_sub(rv.num, rv.num, right.num);
2153 rv = interp_exec(b->left);
2154 right = interp_exec(b->right);
2155 mpq_mul(rv.num, rv.num, right.num);
2158 rv = interp_exec(b->left);
2159 right = interp_exec(b->right);
2160 mpq_div(rv.num, rv.num, right.num);
2163 rv = interp_exec(b->right);
2164 mpq_neg(rv.num, rv.num);
2167 rv = interp_exec(b->right);
2168 mpq_abs(rv.num, rv.num);
2171 rv = interp_exec(b->right);
2174 left = interp_exec(b->left);
2175 right = interp_exec(b->right);
2177 rv.str = text_join(left.str, right.str);
2181 ###### value functions
2183 static struct text text_join(struct text a, struct text b)
2186 rv.len = a.len + b.len;
2187 rv.txt = malloc(rv.len);
2188 memcpy(rv.txt, a.txt, a.len);
2189 memcpy(rv.txt+a.len, b.txt, b.len);
2194 ### Blocks, Statements, and Statement lists.
2196 Now that we have expressions out of the way we need to turn to
2197 statements. There are simple statements and more complex statements.
2198 Simple statements do not contain newlines, complex statements do.
2200 Statements often come in sequences and we have corresponding simple
2201 statement lists and complex statement lists.
2202 The former comprise only simple statements separated by semicolons.
2203 The later comprise complex statements and simple statement lists. They are
2204 separated by newlines. Thus the semicolon is only used to separate
2205 simple statements on the one line. This may be overly restrictive,
2206 but I'm not sure I ever want a complex statement to share a line with
2209 Note that a simple statement list can still use multiple lines if
2210 subsequent lines are indented, so
2212 ###### Example: wrapped simple statement list
2217 is a single simple statement list. This might allow room for
2218 confusion, so I'm not set on it yet.
2220 A simple statement list needs no extra syntax. A complex statement
2221 list has two syntactic forms. It can be enclosed in braces (much like
2222 C blocks), or it can be introduced by a colon and continue until an
2223 unindented newline (much like Python blocks). With this extra syntax
2224 it is referred to as a block.
2226 Note that a block does not have to include any newlines if it only
2227 contains simple statements. So both of:
2229 if condition: a=b; d=f
2231 if condition { a=b; print f }
2235 In either case the list is constructed from a `binode` list with
2236 `Block` as the operator. When parsing the list it is most convenient
2237 to append to the end, so a list is a list and a statement. When using
2238 the list it is more convenient to consider a list to be a statement
2239 and a list. So we need a function to re-order a list.
2240 `reorder_bilist` serves this purpose.
2242 The only stand-alone statement we introduce at this stage is `pass`
2243 which does nothing and is represented as a `NULL` pointer in a `Block`
2244 list. Other stand-alone statements will follow once the infrastructure
2264 Block -> Open Statementlist Close ${ $0 = $<2; }$
2265 | Open Newlines Statementlist Close ${ $0 = $<3; }$
2266 | Open SimpleStatements } ${ $0 = reorder_bilist($<2); }$
2267 | Open Newlines SimpleStatements } ${ $0 = reorder_bilist($<3); }$
2268 | : Statementlist ${ $0 = $<2; }$
2269 | : SimpleStatements ${ $0 = reorder_bilist($<2); }$
2271 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<1); }$
2273 ComplexStatements -> ComplexStatements ComplexStatement ${
2279 | ComplexStatements NEWLINE ${ $0 = $<1; }$
2280 | ComplexStatement ${
2288 ComplexStatement -> SimpleStatements NEWLINE ${
2289 $0 = reorder_bilist($<1);
2291 ## ComplexStatement Grammar
2294 SimpleStatements -> SimpleStatements ; SimpleStatement ${
2300 | SimpleStatement ${
2306 | SimpleStatements ; ${ $0 = $<1; }$
2308 SimpleStatement -> pass ${ $0 = NULL; }$
2309 ## SimpleStatement Grammar
2311 ###### print binode cases
2315 if (b->left == NULL)
2318 print_exec(b->left, indent, 0);
2321 print_exec(b->right, indent, 0);
2324 // block, one per line
2325 if (b->left == NULL)
2326 do_indent(indent, "pass\n");
2328 print_exec(b->left, indent, bracket);
2330 print_exec(b->right, indent, bracket);
2334 ###### propagate binode cases
2337 /* If any statement returns something other than Tnone
2338 * or Tbool then all such must return same type.
2339 * As each statement may be Tnone or something else,
2340 * we must always pass NULL (unknown) down, otherwise an incorrect
2341 * error might occur. We never return Tnone unless it is
2346 for (e = b; e; e = cast(binode, e->right)) {
2347 t = propagate_types(e->left, c, ok, NULL, rules);
2348 if ((rules & Rboolok) && t == Tbool)
2350 if (t && t != Tnone && t != Tbool) {
2353 else if (t != type) {
2354 type_err(c, "error: expected %1%r, found %2",
2355 e->left, type, rules, t);
2363 ###### interp binode cases
2365 while (rv.type == Tnone &&
2368 rv = interp_exec(b->left);
2369 b = cast(binode, b->right);
2373 ### The Print statement
2375 `print` is a simple statement that takes a comma-separated list of
2376 expressions and prints the values separated by spaces and terminated
2377 by a newline. No control of formatting is possible.
2379 `print` faces the same list-ordering issue as blocks, and uses the
2385 ###### SimpleStatement Grammar
2387 | print ExpressionList ${
2388 $0 = reorder_bilist($<2);
2390 | print ExpressionList , ${
2395 $0 = reorder_bilist($0);
2406 ExpressionList -> ExpressionList , Expression ${
2419 ###### print binode cases
2422 do_indent(indent, "print");
2426 print_exec(b->left, -1, 0);
2430 b = cast(binode, b->right);
2436 ###### propagate binode cases
2439 /* don't care but all must be consistent */
2440 propagate_types(b->left, c, ok, NULL, Rnolabel);
2441 propagate_types(b->right, c, ok, NULL, Rnolabel);
2444 ###### interp binode cases
2450 for ( ; b; b = cast(binode, b->right))
2454 left = interp_exec(b->left);
2467 ###### Assignment statement
2469 An assignment will assign a value to a variable, providing it hasn't
2470 be declared as a constant. The analysis phase ensures that the type
2471 will be correct so the interpreter just needs to perform the
2472 calculation. There is a form of assignment which declares a new
2473 variable as well as assigning a value. If a name is assigned before
2474 it is declared, and error will be raised as the name is created as
2475 `Tlabel` and it is illegal to assign to such names.
2481 ###### SimpleStatement Grammar
2482 | Variable = Expression ${ {
2483 struct var *v = cast(var, $1);
2489 if (v->var && v->var->constant) {
2490 type_err(config2context(config), "Cannot assign to a constant: %v",
2491 $0->left, NULL, 0, NULL);
2492 type_err(config2context(config), "name was defined as a constant here",
2493 v->var->where_decl, NULL, 0, NULL);
2496 | VariableDecl = Expression ${
2504 if ($1->var->where_set == NULL) {
2505 type_err(config2context(config), "Variable declared with no type or value: %v",
2515 ###### print binode cases
2518 do_indent(indent, "");
2519 print_exec(b->left, indent, 0);
2521 print_exec(b->right, indent, 0);
2528 struct variable *v = cast(var, b->left)->var;
2529 do_indent(indent, "");
2530 print_exec(b->left, indent, 0);
2531 if (cast(var, b->left)->var->constant) {
2532 if (v->where_decl == v->where_set)
2533 printf("::%.*s ", v->val.type->name.len,
2534 v->val.type->name.txt);
2538 if (v->where_decl == v->where_set)
2539 printf(":%.*s ", v->val.type->name.len,
2540 v->val.type->name.txt);
2546 print_exec(b->right, indent, 0);
2553 ###### propagate binode cases
2557 /* Both must match and not be labels,
2558 * Type must support 'dup',
2559 * result is Tnone */
2560 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
2565 if (propagate_types(b->right, c, ok, t, 0) != t)
2566 if (b->left->type == Xvar)
2567 type_err(c, "info: variable '%v' was set as %1 here.",
2568 cast(var, b->left)->var->where_set, t, rules, Tnone);
2570 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
2572 propagate_types(b->left, c, ok, t, 0);
2574 if (t && t->dup == NULL) {
2575 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
2582 ###### interp binode cases
2586 struct variable *v = cast(var, b->left)->var;
2589 right = interp_exec(b->right);
2598 struct variable *v = cast(var, b->left)->var;
2602 right = interp_exec(b->right);
2604 right = val_init(v->val.type);
2611 ### The `use` statement
2613 The `use` statement is the last "simple" statement. It is needed when
2614 the condition in a conditional statement is a block. `use` works much
2615 like `return` in C, but only completes the `condition`, not the whole
2621 ###### SimpleStatement Grammar
2623 $0 = new_pos(binode, $1);
2628 ###### print binode cases
2631 do_indent(indent, "use ");
2632 print_exec(b->right, -1, 0);
2637 ###### propagate binode cases
2640 /* result matches value */
2641 return propagate_types(b->right, c, ok, type, 0);
2643 ###### interp binode cases
2646 rv = interp_exec(b->right);
2649 ### The Conditional Statement
2651 This is the biggy and currently the only complex statement. This
2652 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
2653 It is comprised of a number of parts, all of which are optional though
2654 set combinations apply. Each part is (usually) a key word (`then` is
2655 sometimes optional) followed by either an expression or a code block,
2656 except the `casepart` which is a "key word and an expression" followed
2657 by a code block. The code-block option is valid for all parts and,
2658 where an expression is also allowed, the code block can use the `use`
2659 statement to report a value. If the code block does not report a value
2660 the effect is similar to reporting `True`.
2662 The `else` and `case` parts, as well as `then` when combined with
2663 `if`, can contain a `use` statement which will apply to some
2664 containing conditional statement. `for` parts, `do` parts and `then`
2665 parts used with `for` can never contain a `use`, except in some
2666 subordinate conditional statement.
2668 If there is a `forpart`, it is executed first, only once.
2669 If there is a `dopart`, then it is executed repeatedly providing
2670 always that the `condpart` or `cond`, if present, does not return a non-True
2671 value. `condpart` can fail to return any value if it simply executes
2672 to completion. This is treated the same as returning `True`.
2674 If there is a `thenpart` it will be executed whenever the `condpart`
2675 or `cond` returns True (or does not return any value), but this will happen
2676 *after* `dopart` (when present).
2678 If `elsepart` is present it will be executed at most once when the
2679 condition returns `False` or some value that isn't `True` and isn't
2680 matched by any `casepart`. If there are any `casepart`s, they will be
2681 executed when the condition returns a matching value.
2683 The particular sorts of values allowed in case parts has not yet been
2684 determined in the language design, so nothing is prohibited.
2686 The various blocks in this complex statement potentially provide scope
2687 for variables as described earlier. Each such block must include the
2688 "OpenScope" nonterminal before parsing the block, and must call
2689 `var_block_close()` when closing the block.
2691 The code following "`if`", "`switch`" and "`for`" does not get its own
2692 scope, but is in a scope covering the whole statement, so names
2693 declared there cannot be redeclared elsewhere. Similarly the
2694 condition following "`while`" is in a scope the covers the body
2695 ("`do`" part) of the loop, and which does not allow conditional scope
2696 extension. Code following "`then`" (both looping and non-looping),
2697 "`else`" and "`case`" each get their own local scope.
2699 The type requirements on the code block in a `whilepart` are quite
2700 unusal. It is allowed to return a value of some identifiable type, in
2701 which case the loop aborts and an appropriate `casepart` is run, or it
2702 can return a Boolean, in which case the loop either continues to the
2703 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
2704 This is different both from the `ifpart` code block which is expected to
2705 return a Boolean, or the `switchpart` code block which is expected to
2706 return the same type as the casepart values. The correct analysis of
2707 the type of the `whilepart` code block is the reason for the
2708 `Rboolok` flag which is passed to `propagate_types()`.
2710 The `cond_statement` cannot fit into a `binode` so a new `exec` is
2719 struct exec *action;
2720 struct casepart *next;
2722 struct cond_statement {
2724 struct exec *forpart, *condpart, *dopart, *thenpart, *elsepart;
2725 struct casepart *casepart;
2728 ###### ast functions
2730 static void free_casepart(struct casepart *cp)
2734 free_exec(cp->value);
2735 free_exec(cp->action);
2742 static void free_cond_statement(struct cond_statement *s)
2746 free_exec(s->forpart);
2747 free_exec(s->condpart);
2748 free_exec(s->dopart);
2749 free_exec(s->thenpart);
2750 free_exec(s->elsepart);
2751 free_casepart(s->casepart);
2755 ###### free exec cases
2756 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
2758 ###### ComplexStatement Grammar
2759 | CondStatement ${ $0 = $<1; }$
2764 // both ForThen and Whilepart open scopes, and CondSuffix only
2765 // closes one - so in the first branch here we have another to close.
2766 CondStatement -> ForThen WhilePart CondSuffix ${
2768 $0->forpart = $1.forpart; $1.forpart = NULL;
2769 $0->thenpart = $1.thenpart; $1.thenpart = NULL;
2770 $0->condpart = $2.condpart; $2.condpart = NULL;
2771 $0->dopart = $2.dopart; $2.dopart = NULL;
2772 var_block_close(config2context(config), CloseSequential);
2774 | WhilePart CondSuffix ${
2776 $0->condpart = $1.condpart; $1.condpart = NULL;
2777 $0->dopart = $1.dopart; $1.dopart = NULL;
2779 | SwitchPart CondSuffix ${
2783 | IfPart IfSuffix ${
2785 $0->condpart = $1.condpart; $1.condpart = NULL;
2786 $0->thenpart = $1.thenpart; $1.thenpart = NULL;
2787 // This is where we close an "if" statement
2788 var_block_close(config2context(config), CloseSequential);
2791 CondSuffix -> IfSuffix ${
2793 // This is where we close scope of the whole
2794 // "for" or "while" statement
2795 var_block_close(config2context(config), CloseSequential);
2797 | CasePart CondSuffix ${
2799 $1->next = $0->casepart;
2804 CasePart -> Newlines case Expression OpenScope Block ${
2805 $0 = calloc(1,sizeof(struct casepart));
2808 var_block_close(config2context(config), CloseParallel);
2810 | case Expression OpenScope Block ${
2811 $0 = calloc(1,sizeof(struct casepart));
2814 var_block_close(config2context(config), CloseParallel);
2818 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
2819 | Newlines else OpenScope Block ${
2820 $0 = new(cond_statement);
2822 var_block_close(config2context(config), CloseElse);
2824 | else OpenScope Block ${
2825 $0 = new(cond_statement);
2827 var_block_close(config2context(config), CloseElse);
2829 | Newlines else OpenScope CondStatement ${
2830 $0 = new(cond_statement);
2832 var_block_close(config2context(config), CloseElse);
2834 | else OpenScope CondStatement ${
2835 $0 = new(cond_statement);
2837 var_block_close(config2context(config), CloseElse);
2842 // These scopes are closed in CondSuffix
2843 ForPart -> for OpenScope SimpleStatements ${
2844 $0 = reorder_bilist($<3);
2846 | for OpenScope Block ${
2850 ThenPart -> then OpenScope SimpleStatements ${
2851 $0 = reorder_bilist($<3);
2852 var_block_close(config2context(config), CloseSequential);
2854 | then OpenScope Block ${
2856 var_block_close(config2context(config), CloseSequential);
2859 ThenPartNL -> ThenPart OptNL ${
2863 // This scope is closed in CondSuffix
2864 WhileHead -> while OpenScope Block ${
2869 ForThen -> ForPart OptNL ThenPartNL ${
2877 // This scope is closed in CondSuffix
2878 WhilePart -> while OpenScope Expression Block ${
2879 $0.type = Xcond_statement;
2883 | WhileHead OptNL do Block ${
2884 $0.type = Xcond_statement;
2889 IfPart -> if OpenScope Expression OpenScope Block ${
2890 $0.type = Xcond_statement;
2893 var_block_close(config2context(config), CloseParallel);
2895 | if OpenScope Block OptNL then OpenScope Block ${
2896 $0.type = Xcond_statement;
2899 var_block_close(config2context(config), CloseParallel);
2903 // This scope is closed in CondSuffix
2904 SwitchPart -> switch OpenScope Expression ${
2907 | switch OpenScope Block ${
2911 ###### print exec cases
2913 case Xcond_statement:
2915 struct cond_statement *cs = cast(cond_statement, e);
2916 struct casepart *cp;
2918 do_indent(indent, "for");
2919 if (bracket) printf(" {\n"); else printf(":\n");
2920 print_exec(cs->forpart, indent+1, bracket);
2923 do_indent(indent, "} then {\n");
2925 do_indent(indent, "then:\n");
2926 print_exec(cs->thenpart, indent+1, bracket);
2928 if (bracket) do_indent(indent, "}\n");
2932 if (cs->condpart && cs->condpart->type == Xbinode &&
2933 cast(binode, cs->condpart)->op == Block) {
2935 do_indent(indent, "while {\n");
2937 do_indent(indent, "while:\n");
2938 print_exec(cs->condpart, indent+1, bracket);
2940 do_indent(indent, "} do {\n");
2942 do_indent(indent, "do:\n");
2943 print_exec(cs->dopart, indent+1, bracket);
2945 do_indent(indent, "}\n");
2947 do_indent(indent, "while ");
2948 print_exec(cs->condpart, 0, bracket);
2953 print_exec(cs->dopart, indent+1, bracket);
2955 do_indent(indent, "}\n");
2960 do_indent(indent, "switch");
2962 do_indent(indent, "if");
2963 if (cs->condpart && cs->condpart->type == Xbinode &&
2964 cast(binode, cs->condpart)->op == Block) {
2969 print_exec(cs->condpart, indent+1, bracket);
2971 do_indent(indent, "}\n");
2973 do_indent(indent, "then:\n");
2974 print_exec(cs->thenpart, indent+1, bracket);
2978 print_exec(cs->condpart, 0, bracket);
2984 print_exec(cs->thenpart, indent+1, bracket);
2986 do_indent(indent, "}\n");
2991 for (cp = cs->casepart; cp; cp = cp->next) {
2992 do_indent(indent, "case ");
2993 print_exec(cp->value, -1, 0);
2998 print_exec(cp->action, indent+1, bracket);
3000 do_indent(indent, "}\n");
3003 do_indent(indent, "else");
3008 print_exec(cs->elsepart, indent+1, bracket);
3010 do_indent(indent, "}\n");
3015 ###### propagate exec cases
3016 case Xcond_statement:
3018 // forpart and dopart must return Tnone
3019 // thenpart must return Tnone if there is a dopart,
3020 // otherwise it is like elsepart.
3022 // be bool if there is no casepart
3023 // match casepart->values if there is a switchpart
3024 // either be bool or match casepart->value if there
3026 // elsepart and casepart->action must match the return type
3027 // expected of this statement.
3028 struct cond_statement *cs = cast(cond_statement, prog);
3029 struct casepart *cp;
3031 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
3032 if (!vtype_compat(Tnone, t, 0))
3034 t = propagate_types(cs->dopart, c, ok, Tnone, 0);
3035 if (!vtype_compat(Tnone, t, 0))
3038 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
3039 if (!vtype_compat(Tnone, t, 0))
3042 if (cs->casepart == NULL)
3043 propagate_types(cs->condpart, c, ok, Tbool, 0);
3045 /* Condpart must match case values, with bool permitted */
3047 for (cp = cs->casepart;
3048 cp && !t; cp = cp->next)
3049 t = propagate_types(cp->value, c, ok, NULL, 0);
3050 if (!t && cs->condpart)
3051 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok);
3052 // Now we have a type (I hope) push it down
3054 for (cp = cs->casepart; cp; cp = cp->next)
3055 propagate_types(cp->value, c, ok, t, 0);
3056 propagate_types(cs->condpart, c, ok, t, Rboolok);
3059 // (if)then, else, and case parts must return expected type.
3060 if (!cs->dopart && !type)
3061 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
3063 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
3064 for (cp = cs->casepart;
3067 type = propagate_types(cp->action, c, ok, NULL, rules);
3070 propagate_types(cs->thenpart, c, ok, type, rules);
3071 propagate_types(cs->elsepart, c, ok, type, rules);
3072 for (cp = cs->casepart; cp ; cp = cp->next)
3073 propagate_types(cp->action, c, ok, type, rules);
3079 ###### interp exec cases
3080 case Xcond_statement:
3082 struct value v, cnd;
3083 struct casepart *cp;
3084 struct cond_statement *c = cast(cond_statement, e);
3087 interp_exec(c->forpart);
3090 cnd = interp_exec(c->condpart);
3093 if (!(cnd.type == Tnone ||
3094 (cnd.type == Tbool && cnd.bool != 0)))
3096 // cnd is Tnone or Tbool, doesn't need to be freed
3098 interp_exec(c->dopart);
3101 v = interp_exec(c->thenpart);
3102 if (v.type != Tnone || !c->dopart)
3106 } while (c->dopart);
3108 for (cp = c->casepart; cp; cp = cp->next) {
3109 v = interp_exec(cp->value);
3110 if (value_cmp(v, cnd) == 0) {
3113 return interp_exec(cp->action);
3119 return interp_exec(c->elsepart);
3124 ### Finally the whole program.
3126 Somewhat reminiscent of Pascal a (current) Ocean program starts with
3127 the keyword "program" and a list of variable names which are assigned
3128 values from command line arguments. Following this is a `block` which
3129 is the code to execute.
3131 As this is the top level, several things are handled a bit
3133 The whole program is not interpreted by `interp_exec` as that isn't
3134 passed the argument list which the program requires. Similarly type
3135 analysis is a bit more interesting at this level.
3140 ###### Parser: grammar
3143 Program -> program OpenScope Varlist Block OptNL ${
3146 $0->left = reorder_bilist($<3);
3148 var_block_close(config2context(config), CloseSequential);
3149 if (config2context(config)->scope_stack) abort();
3152 tok_err(config2context(config),
3153 "error: unhandled parse error", &$1);
3156 Varlist -> Varlist ArgDecl ${
3165 ArgDecl -> IDENTIFIER ${ {
3166 struct variable *v = var_decl(config2context(config), $1.txt);
3173 ###### print binode cases
3175 do_indent(indent, "program");
3176 for (b2 = cast(binode, b->left); b2; b2 = cast(binode, b2->right)) {
3178 print_exec(b2->left, 0, 0);
3184 print_exec(b->right, indent+1, bracket);
3186 do_indent(indent, "}\n");
3189 ###### propagate binode cases
3190 case Program: abort();
3192 ###### core functions
3194 static int analyse_prog(struct exec *prog, struct parse_context *c)
3196 struct binode *b = cast(binode, prog);
3203 propagate_types(b->right, c, &ok, Tnone, 0);
3208 for (b = cast(binode, b->left); b; b = cast(binode, b->right)) {
3209 struct var *v = cast(var, b->left);
3210 if (!v->var->val.type) {
3211 v->var->where_set = b;
3212 v->var->val = val_prepare(Tstr);
3215 b = cast(binode, prog);
3218 propagate_types(b->right, c, &ok, Tnone, 0);
3223 /* Make sure everything is still consistent */
3224 propagate_types(b->right, c, &ok, Tnone, 0);
3228 static void interp_prog(struct exec *prog, char **argv)
3230 struct binode *p = cast(binode, prog);
3236 al = cast(binode, p->left);
3238 struct var *v = cast(var, al->left);
3239 struct value *vl = &v->var->val;
3241 if (argv[0] == NULL) {
3242 printf("Not enough args\n");
3245 al = cast(binode, al->right);
3247 *vl = parse_value(vl->type, argv[0]);
3248 if (vl->type == NULL)
3252 v = interp_exec(p->right);
3256 ###### interp binode cases
3257 case Program: abort();
3259 ## And now to test it out.
3261 Having a language requires having a "hello world" program. I'll
3262 provide a little more than that: a program that prints "Hello world"
3263 finds the GCD of two numbers, prints the first few elements of
3264 Fibonacci, and performs a binary search for a number.
3266 ###### File: oceani.mk
3269 @echo "===== TEST ====="
3270 ./oceani --section "test: hello" oceani.mdc 55 33
3275 print "Hello World, what lovely oceans you have!"
3276 /* When a variable is defined in both branches of an 'if',
3277 * and used afterwards, the variables are merged.
3283 print "Is", A, "bigger than", B,"? ", bigger
3284 /* If a variable is not used after the 'if', no
3285 * merge happens, so types can be different
3288 double:string = "yes"
3289 print A, "is more than twice", B, "?", double
3292 print "double", A, "is only", double
3303 print "GCD of", A, "and", B,"is", a
3305 print a, "is not positive, cannot calculate GCD"
3307 print b, "is not positive, cannot calculate GCD"
3312 print "Fibonacci:", f1,f2,
3313 then togo = togo - 1
3321 /* Binary search... */
3326 mid := (lo + hi) / 2
3338 print "Yay, I found", target
3340 print "Closest I found was", mid