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 Elements that are present purely to make a usable language, and
45 without any expectation that they will remain, are the "program'
46 clause, which provides a list of variables to received command-line
47 arguments, and the "print" statement which performs simple output.
49 The current scalar types are "number", "Boolean", and "string".
50 Boolean will likely stay in its current form, the other two might, but
51 could just as easily be changed.
55 Versions of the interpreter which obviously do not support a complete
56 language will be named after creeks and streams. This one is Jamison
59 Once we have something reasonably resembling a complete language, the
60 names of rivers will be used.
61 Early versions of the compiler will be named after seas. Major
62 releases of the compiler will be named after oceans. Hopefully I will
63 be finished once I get to the Pacific Ocean release.
67 As well as parsing and executing a program, the interpreter can print
68 out the program from the parsed internal structure. This is useful
69 for validating the parsing.
70 So the main requirements of the interpreter are:
72 - Parse the program, possibly with tracing,
73 - Analyse the parsed program to ensure consistency,
75 - Execute the program, if no parsing or consistency errors were found.
77 This is all performed by a single C program extracted with
80 There will be two formats for printing the program: a default and one
81 that uses bracketing. So a `--bracket` command line option is needed
82 for that. Normally the first code section found is used, however an
83 alternate section can be requested so that a file (such as this one)
84 can contain multiple programs. This is effected with the `--section`
87 This code must be compiled with `-fplan9-extensions` so that anonymous
88 structures can be used.
90 ###### File: oceani.mk
92 myCFLAGS := -Wall -g -fplan9-extensions
93 CFLAGS := $(filter-out $(myCFLAGS),$(CFLAGS)) $(myCFLAGS)
94 myLDLIBS:= libparser.o libscanner.o libmdcode.o -licuuc
95 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
97 all :: $(LDLIBS) oceani
98 oceani.c oceani.h : oceani.mdc parsergen
99 ./parsergen -o oceani --LALR --tag Parser oceani.mdc
100 oceani.mk: oceani.mdc md2c
103 oceani: oceani.o $(LDLIBS)
104 $(CC) $(CFLAGS) -o oceani oceani.o $(LDLIBS)
106 ###### Parser: header
109 struct parse_context {
110 struct token_config config;
119 #define container_of(ptr, type, member) ({ \
120 const typeof( ((type *)0)->member ) *__mptr = (ptr); \
121 (type *)( (char *)__mptr - offsetof(type,member) );})
123 #define config2context(_conf) container_of(_conf, struct parse_context, \
126 ###### Parser: reduce
127 struct parse_context *c = config2context(config);
135 #include <sys/mman.h>
154 static char Usage[] =
155 "Usage: oceani --trace --print --noexec --brackets --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[])
170 struct section *s, *ss;
171 char *section = NULL;
172 struct parse_context context = {
174 .ignored = (1 << TK_mark),
175 .number_chars = ".,_+- ",
180 int doprint=0, dotrace=0, doexec=1, brackets=0;
182 while ((opt = getopt_long(argc, argv, options, long_options, NULL))
185 case 't': dotrace=1; break;
186 case 'p': doprint=1; break;
187 case 'n': doexec=0; break;
188 case 'b': brackets=1; break;
189 case 's': section = optarg; break;
190 default: fprintf(stderr, Usage);
194 if (optind >= argc) {
195 fprintf(stderr, "oceani: no input file given\n");
198 fd = open(argv[optind], O_RDONLY);
200 fprintf(stderr, "oceani: cannot open %s\n", argv[optind]);
203 context.file_name = argv[optind];
204 len = lseek(fd, 0, 2);
205 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
206 s = code_extract(file, file+len, NULL);
208 fprintf(stderr, "oceani: could not find any code in %s\n",
213 ## context initialization
216 for (ss = s; ss; ss = ss->next) {
217 struct text sec = ss->section;
218 if (sec.len == strlen(section) &&
219 strncmp(sec.txt, section, sec.len) == 0)
223 fprintf(stderr, "oceani: cannot find section %s\n",
229 parse_oceani(ss->code, &context.config, dotrace ? stderr : NULL);
232 fprintf(stderr, "oceani: no program found.\n");
233 context.parse_error = 1;
235 if (context.prog && doprint) {
238 print_exec(context.prog, 0, brackets);
240 if (context.prog && doexec && !context.parse_error) {
241 if (!analyse_prog(context.prog, &context)) {
242 fprintf(stderr, "oceani: type error in program - not running.\n");
245 interp_prog(context.prog, argv+optind+1);
247 free_exec(context.prog);
250 struct section *t = s->next;
256 ## free context types
257 exit(context.parse_error ? 1 : 0);
262 The four requirements of parse, analyse, print, interpret apply to
263 each language element individually so that is how most of the code
266 Three of the four are fairly self explanatory. The one that requires
267 a little explanation is the analysis step.
269 The current language design does not require the types of variables to
270 be declared, but they must still have a single type. Different
271 operations impose different requirements on the variables, for example
272 addition requires both arguments to be numeric, and assignment
273 requires the variable on the left to have the same type as the
274 expression on the right.
276 Analysis involves propagating these type requirements around and
277 consequently setting the type of each variable. If any requirements
278 are violated (e.g. a string is compared with a number) or if a
279 variable needs to have two different types, then an error is raised
280 and the program will not run.
282 If the same variable is declared in both branchs of an 'if/else', or
283 in all cases of a 'switch' then the multiple instances may be merged
284 into just one variable if the variable is referenced after the
285 conditional statement. When this happens, the types must naturally be
286 consistent across all the branches. When the variable is not used
287 outside the if, the variables in the different branches are distinct
288 and can be of different types.
290 Undeclared names may only appear in "use" statements and "case" expressions.
291 These names are given a type of "label" and a unique value.
292 This allows them to fill the role of a name in an enumerated type, which
293 is useful for testing the `switch` statement.
295 As we will see, the condition part of a `while` statement can return
296 either a Boolean or some other type. This requires that the expected
297 type that gets passed around comprises a type and a flag to indicate
298 that `Tbool` is also permitted.
300 As there are, as yet, no distinct types that are compatible, there
301 isn't much subtlety in the analysis. When we have distinct number
302 types, this will become more interesting.
306 When analysis discovers an inconsistency it needs to report an error;
307 just refusing to run the code ensures that the error doesn't cascade,
308 but by itself it isn't very useful. A clear understanding of the sort
309 of error message that are useful will help guide the process of
312 At a simplistic level, the only sort of error that type analysis can
313 report is that the type of some construct doesn't match a contextual
314 requirement. For example, in `4 + "hello"` the addition provides a
315 contextual requirement for numbers, but `"hello"` is not a number. In
316 this particular example no further information is needed as the types
317 are obvious from local information. When a variable is involved that
318 isn't the case. It may be helpful to explain why the variable has a
319 particular type, by indicating the location where the type was set,
320 whether by declaration or usage.
322 Using a recursive-descent analysis we can easily detect a problem at
323 multiple locations. In "`hello:= "there"; 4 + hello`" the addition
324 will detect that one argument is not a number and the usage of `hello`
325 will detect that a number was wanted, but not provided. In this
326 (early) version of the language, we will generate error reports at
327 multiple locations, so the use of `hello` will report an error and
328 explain were the value was set, and the addition will report an error
329 and say why numbers are needed. To be able to report locations for
330 errors, each language element will need to record a file location
331 (line and column) and each variable will need to record the language
332 element where its type was set. For now we will assume that each line
333 of an error message indicates one location in the file, and up to 2
334 types. So we provide a `printf`-like function which takes a format, a
335 location (a `struct exec` which has not yet been introduced), and 2
336 types. "`%1`" reports the first type, "`%2`" reports the second. We
337 will need a function to print the location, once we know how that is
338 stored. e As will be explained later, there are sometimes extra rules for
339 type matching and they might affect error messages, we need to pass those
342 As well as type errors, we sometimes need to report problems with
343 tokens, which might be unexpected or might name a type that has not
344 been defined. For these we have `tok_err()` which reports an error
345 with a given token. Each of the error functions sets the flag in the
346 context so indicate that parsing failed.
350 static void fput_loc(struct exec *loc, FILE *f);
352 ###### core functions
354 static void type_err(struct parse_context *c,
355 char *fmt, struct exec *loc,
356 struct type *t1, int rules, struct type *t2)
358 fprintf(stderr, "%s:", c->file_name);
359 fput_loc(loc, stderr);
360 for (; *fmt ; fmt++) {
367 case '%': fputc(*fmt, stderr); break; // NOTEST
368 default: fputc('?', stderr); break; // NOTEST
370 type_print(t1, stderr);
373 type_print(t2, stderr);
382 static void tok_err(struct parse_context *c, char *fmt, struct token *t)
384 fprintf(stderr, "%s:%d:%d: %s: %.*s\n", c->file_name, t->line, t->col, fmt,
385 t->txt.len, t->txt.txt);
389 ## Entities: declared and predeclared.
391 There are various "things" that the language and/or the interpreter
392 needs to know about to parse and execute a program. These include
393 types, variables, values, and executable code. These are all lumped
394 together under the term "entities" (calling them "objects" would be
395 confusing) and introduced here. The following section will present the
396 different specific code elements which comprise or manipulate these
401 Values come in a wide range of types, with more likely to be added.
402 Each type needs to be able to print its own values (for convenience at
403 least) as well as to compare two values, at least for equality and
404 possibly for order. For now, values might need to be duplicated and
405 freed, though eventually such manipulations will be better integrated
408 Rather than requiring every numeric type to support all numeric
409 operations (add, multiple, etc), we allow types to be able to present
410 as one of a few standard types: integer, float, and fraction. The
411 existence of these conversion functions eventually enable types to
412 determine if they are compatible with other types, though such types
413 have not yet been implemented.
415 Named type are stored in a simple linked list. Objects of each type are
416 "values" which are often passed around by value.
423 ## value union fields
431 void (*init)(struct type *type, struct value *val);
432 void (*print)(struct type *type, struct value *val);
433 void (*print_type)(struct type *type, FILE *f);
434 int (*cmp_order)(struct type *t1, struct type *t2,
435 struct value *v1, struct value *v2);
436 int (*cmp_eq)(struct type *t1, struct type *t2,
437 struct value *v1, struct value *v2);
438 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
439 void (*free)(struct type *type, struct value *val);
440 void (*free_type)(struct type *t);
441 long long (*to_int)(struct value *v);
442 double (*to_float)(struct value *v);
443 int (*to_mpq)(mpq_t *q, struct value *v);
452 struct type *typelist;
456 static struct type *find_type(struct parse_context *c, struct text s)
458 struct type *l = c->typelist;
461 text_cmp(l->name, s) != 0)
466 static struct type *add_type(struct parse_context *c, struct text s,
471 n = calloc(1, sizeof(*n));
474 n->next = c->typelist;
479 static void free_type(struct type *t)
481 /* The type is always a reference to something in the
482 * context, so we don't need to free anything.
486 static void free_value(struct type *type, struct value *v)
492 static void type_print(struct type *type, FILE *f)
495 fputs("*unknown*type*", f);
496 else if (type->name.len)
497 fprintf(f, "%.*s", type->name.len, type->name.txt);
498 else if (type->print_type)
499 type->print_type(type, f);
501 fputs("*invalid*type*", f); // NOTEST
504 static void val_init(struct type *type, struct value *val)
506 if (type && type->init)
507 type->init(type, val);
510 static void dup_value(struct type *type,
511 struct value *vold, struct value *vnew)
513 if (type && type->dup)
514 type->dup(type, vold, vnew);
517 static int value_cmp(struct type *tl, struct type *tr,
518 struct value *left, struct value *right)
520 if (tl && tl->cmp_order)
521 return tl->cmp_order(tl, tr, left, right);
522 if (tl && tl->cmp_eq)
523 return tl->cmp_eq(tl, tr, left, right);
527 static void print_value(struct type *type, struct value *v)
529 if (type && type->print)
530 type->print(type, v);
532 printf("*Unknown*"); // NOTEST
535 static struct value *val_alloc(struct type *t, struct value *init)
541 ret = calloc(1, t->size);
543 memcpy(ret, init, t->size);
551 static void free_value(struct type *type, struct value *v);
552 static int type_compat(struct type *require, struct type *have, int rules);
553 static void type_print(struct type *type, FILE *f);
554 static void val_init(struct type *type, struct value *v);
555 static void dup_value(struct type *type,
556 struct value *vold, struct value *vnew);
557 static int value_cmp(struct type *tl, struct type *tr,
558 struct value *left, struct value *right);
559 static void print_value(struct type *type, struct value *v);
561 ###### free context types
563 while (context.typelist) {
564 struct type *t = context.typelist;
566 context.typelist = t->next;
574 Values of the base types can be numbers, which we represent as
575 multi-precision fractions, strings, Booleans and labels. When
576 analysing the program we also need to allow for places where no value
577 is meaningful (type `Tnone`) and where we don't know what type to
578 expect yet (type is `NULL`).
580 Values are never shared, they are always copied when used, and freed
581 when no longer needed.
583 When propagating type information around the program, we need to
584 determine if two types are compatible, where type `NULL` is compatible
585 with anything. There are two special cases with type compatibility,
586 both related to the Conditional Statement which will be described
587 later. In some cases a Boolean can be accepted as well as some other
588 primary type, and in others any type is acceptable except a label (`Vlabel`).
589 A separate function encoding these cases will simplify some code later.
593 int (*compat)(struct type *this, struct type *other);
597 static int type_compat(struct type *require, struct type *have, int rules)
599 if ((rules & Rboolok) && have == Tbool)
601 if ((rules & Rnolabel) && have == Tlabel)
603 if (!require || !have)
607 return require->compat(require, have);
609 return require == have;
614 #include "parse_string.h"
615 #include "parse_number.h"
618 myLDLIBS := libnumber.o libstring.o -lgmp
619 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
621 ###### type union fields
622 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
624 ###### value union fields
631 static void _free_value(struct type *type, struct value *v)
635 switch (type->vtype) {
637 case Vstr: free(v->str.txt); break;
638 case Vnum: mpq_clear(v->num); break;
644 ###### value functions
646 static void _val_init(struct type *type, struct value *val)
648 switch(type->vtype) {
649 case Vnone: // NOTEST
652 mpq_init(val->num); break;
654 val->str.txt = malloc(1);
660 case Vlabel: // NOTEST
661 val->label = NULL; // NOTEST
666 static void _dup_value(struct type *type,
667 struct value *vold, struct value *vnew)
669 switch (type->vtype) {
670 case Vnone: // NOTEST
673 vnew->label = vold->label;
676 vnew->bool = vold->bool;
680 mpq_set(vnew->num, vold->num);
683 vnew->str.len = vold->str.len;
684 vnew->str.txt = malloc(vnew->str.len);
685 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
690 static int _value_cmp(struct type *tl, struct type *tr,
691 struct value *left, struct value *right)
695 return tl - tr; // NOTEST
697 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
698 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
699 case Vstr: cmp = text_cmp(left->str, right->str); break;
700 case Vbool: cmp = left->bool - right->bool; break;
701 case Vnone: cmp = 0; // NOTEST
706 static void _print_value(struct type *type, struct value *v)
708 switch (type->vtype) {
709 case Vnone: // NOTEST
710 printf("*no-value*"); break; // NOTEST
711 case Vlabel: // NOTEST
712 printf("*label-%p*", v->label); break; // NOTEST
714 printf("%.*s", v->str.len, v->str.txt); break;
716 printf("%s", v->bool ? "True":"False"); break;
721 mpf_set_q(fl, v->num);
722 gmp_printf("%Fg", fl);
729 static void _free_value(struct type *type, struct value *v);
731 static struct type base_prototype = {
733 .print = _print_value,
734 .cmp_order = _value_cmp,
735 .cmp_eq = _value_cmp,
740 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
743 static struct type *add_base_type(struct parse_context *c, char *n,
744 enum vtype vt, int size)
746 struct text txt = { n, strlen(n) };
749 t = add_type(c, txt, &base_prototype);
752 t->align = size > sizeof(void*) ? sizeof(void*) : size;
753 if (t->size & (t->align - 1))
754 t->size = (t->size | (t->align - 1)) + 1;
758 ###### context initialization
760 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
761 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
762 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
763 Tnone = add_base_type(&context, "none", Vnone, 0);
764 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
768 Variables are scoped named values. We store the names in a linked list
769 of "bindings" sorted in lexical order, and use sequential search and
776 struct binding *next; // in lexical order
780 This linked list is stored in the parse context so that "reduce"
781 functions can find or add variables, and so the analysis phase can
782 ensure that every variable gets a type.
786 struct binding *varlist; // In lexical order
790 static struct binding *find_binding(struct parse_context *c, struct text s)
792 struct binding **l = &c->varlist;
797 (cmp = text_cmp((*l)->name, s)) < 0)
801 n = calloc(1, sizeof(*n));
808 Each name can be linked to multiple variables defined in different
809 scopes. Each scope starts where the name is declared and continues
810 until the end of the containing code block. Scopes of a given name
811 cannot nest, so a declaration while a name is in-scope is an error.
813 ###### binding fields
814 struct variable *var;
818 struct variable *previous;
821 struct binding *name;
822 struct exec *where_decl;// where name was declared
823 struct exec *where_set; // where type was set
827 While the naming seems strange, we include local constants in the
828 definition of variables. A name declared `var := value` can
829 subsequently be changed, but a name declared `var ::= value` cannot -
832 ###### variable fields
835 Scopes in parallel branches can be partially merged. More
836 specifically, if a given name is declared in both branches of an
837 if/else then its scope is a candidate for merging. Similarly if
838 every branch of an exhaustive switch (e.g. has an "else" clause)
839 declares a given name, then the scopes from the branches are
840 candidates for merging.
842 Note that names declared inside a loop (which is only parallel to
843 itself) are never visible after the loop. Similarly names defined in
844 scopes which are not parallel, such as those started by `for` and
845 `switch`, are never visible after the scope. Only variables defined in
846 both `then` and `else` (including the implicit then after an `if`, and
847 excluding `then` used with `for`) and in all `case`s and `else` of a
848 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
850 Labels, which are a bit like variables, follow different rules.
851 Labels are not explicitly declared, but if an undeclared name appears
852 in a context where a label is legal, that effectively declares the
853 name as a label. The declaration remains in force (or in scope) at
854 least to the end of the immediately containing block and conditionally
855 in any larger containing block which does not declare the name in some
856 other way. Importantly, the conditional scope extension happens even
857 if the label is only used in one parallel branch of a conditional --
858 when used in one branch it is treated as having been declared in all
861 Merge candidates are tentatively visible beyond the end of the
862 branching statement which creates them. If the name is used, the
863 merge is affirmed and they become a single variable visible at the
864 outer layer. If not - if it is redeclared first - the merge lapses.
866 To track scopes we have an extra stack, implemented as a linked list,
867 which roughly parallels the parse stack and which is used exclusively
868 for scoping. When a new scope is opened, a new frame is pushed and
869 the child-count of the parent frame is incremented. This child-count
870 is used to distinguish between the first of a set of parallel scopes,
871 in which declared variables must not be in scope, and subsequent
872 branches, whether they may already be conditionally scoped.
874 To push a new frame *before* any code in the frame is parsed, we need a
875 grammar reduction. This is most easily achieved with a grammar
876 element which derives the empty string, and creates the new scope when
877 it is recognised. This can be placed, for example, between a keyword
878 like "if" and the code following it.
882 struct scope *parent;
888 struct scope *scope_stack;
891 static void scope_pop(struct parse_context *c)
893 struct scope *s = c->scope_stack;
895 c->scope_stack = s->parent;
900 static void scope_push(struct parse_context *c)
902 struct scope *s = calloc(1, sizeof(*s));
904 c->scope_stack->child_count += 1;
905 s->parent = c->scope_stack;
913 OpenScope -> ${ scope_push(c); }$
914 ClosePara -> ${ var_block_close(c, CloseParallel); }$
916 Each variable records a scope depth and is in one of four states:
918 - "in scope". This is the case between the declaration of the
919 variable and the end of the containing block, and also between
920 the usage with affirms a merge and the end of that block.
922 The scope depth is not greater than the current parse context scope
923 nest depth. When the block of that depth closes, the state will
924 change. To achieve this, all "in scope" variables are linked
925 together as a stack in nesting order.
927 - "pending". The "in scope" block has closed, but other parallel
928 scopes are still being processed. So far, every parallel block at
929 the same level that has closed has declared the name.
931 The scope depth is the depth of the last parallel block that
932 enclosed the declaration, and that has closed.
934 - "conditionally in scope". The "in scope" block and all parallel
935 scopes have closed, and no further mention of the name has been
936 seen. This state includes a secondary nest depth which records the
937 outermost scope seen since the variable became conditionally in
938 scope. If a use of the name is found, the variable becomes "in
939 scope" and that secondary depth becomes the recorded scope depth.
940 If the name is declared as a new variable, the old variable becomes
941 "out of scope" and the recorded scope depth stays unchanged.
943 - "out of scope". The variable is neither in scope nor conditionally
944 in scope. It is permanently out of scope now and can be removed from
945 the "in scope" stack.
947 ###### variable fields
948 int depth, min_depth;
949 enum { OutScope, PendingScope, CondScope, InScope } scope;
950 struct variable *in_scope;
954 struct variable *in_scope;
956 All variables with the same name are linked together using the
957 'previous' link. Those variable that have been affirmatively merged all
958 have a 'merged' pointer that points to one primary variable - the most
959 recently declared instance. When merging variables, we need to also
960 adjust the 'merged' pointer on any other variables that had previously
961 been merged with the one that will no longer be primary.
963 A variable that is no longer the most recent instance of a name may
964 still have "pending" scope, if it might still be merged with most
965 recent instance. These variables don't really belong in the
966 "in_scope" list, but are not immediately removed when a new instance
967 is found. Instead, they are detected and ignored when considering the
968 list of in_scope names.
970 ###### variable fields
971 struct variable *merged;
975 static void variable_merge(struct variable *primary, struct variable *secondary)
981 primary = primary->merged;
983 for (v = primary->previous; v; v=v->previous)
984 if (v == secondary || v == secondary->merged ||
985 v->merged == secondary ||
986 (v->merged && v->merged == secondary->merged)) {
992 ###### free context vars
994 while (context.varlist) {
995 struct binding *b = context.varlist;
996 struct variable *v = b->var;
997 context.varlist = b->next;
1000 struct variable *t = v;
1003 free_value(t->type, t->val);
1006 // This is a global constant
1007 free_exec(t->where_decl);
1012 #### Manipulating Bindings
1014 When a name is conditionally visible, a new declaration discards the
1015 old binding - the condition lapses. Conversely a usage of the name
1016 affirms the visibility and extends it to the end of the containing
1017 block - i.e. the block that contains both the original declaration and
1018 the latest usage. This is determined from `min_depth`. When a
1019 conditionally visible variable gets affirmed like this, it is also
1020 merged with other conditionally visible variables with the same name.
1022 When we parse a variable declaration we either report an error if the
1023 name is currently bound, or create a new variable at the current nest
1024 depth if the name is unbound or bound to a conditionally scoped or
1025 pending-scope variable. If the previous variable was conditionally
1026 scoped, it and its homonyms becomes out-of-scope.
1028 When we parse a variable reference (including non-declarative assignment
1029 "foo = bar") we report an error if the name is not bound or is bound to
1030 a pending-scope variable; update the scope if the name is bound to a
1031 conditionally scoped variable; or just proceed normally if the named
1032 variable is in scope.
1034 When we exit a scope, any variables bound at this level are either
1035 marked out of scope or pending-scoped, depending on whether the scope
1036 was sequential or parallel. Here a "parallel" scope means the "then"
1037 or "else" part of a conditional, or any "case" or "else" branch of a
1038 switch. Other scopes are "sequential".
1040 When exiting a parallel scope we check if there are any variables that
1041 were previously pending and are still visible. If there are, then
1042 there weren't redeclared in the most recent scope, so they cannot be
1043 merged and must become out-of-scope. If it is not the first of
1044 parallel scopes (based on `child_count`), we check that there was a
1045 previous binding that is still pending-scope. If there isn't, the new
1046 variable must now be out-of-scope.
1048 When exiting a sequential scope that immediately enclosed parallel
1049 scopes, we need to resolve any pending-scope variables. If there was
1050 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1051 we need to mark all pending-scope variable as out-of-scope. Otherwise
1052 all pending-scope variables become conditionally scoped.
1055 enum closetype { CloseSequential, CloseParallel, CloseElse };
1057 ###### ast functions
1059 static struct variable *var_decl(struct parse_context *c, struct text s)
1061 struct binding *b = find_binding(c, s);
1062 struct variable *v = b->var;
1064 switch (v ? v->scope : OutScope) {
1066 /* Caller will report the error */
1070 v && v->scope == CondScope;
1072 v->scope = OutScope;
1076 v = calloc(1, sizeof(*v));
1077 v->previous = b->var;
1080 v->min_depth = v->depth = c->scope_depth;
1082 v->in_scope = c->in_scope;
1088 static struct variable *var_ref(struct parse_context *c, struct text s)
1090 struct binding *b = find_binding(c, s);
1091 struct variable *v = b->var;
1092 struct variable *v2;
1094 switch (v ? v->scope : OutScope) {
1097 /* Caller will report the error */
1100 /* All CondScope variables of this name need to be merged
1101 * and become InScope
1103 v->depth = v->min_depth;
1105 for (v2 = v->previous;
1106 v2 && v2->scope == CondScope;
1108 variable_merge(v, v2);
1116 static void var_block_close(struct parse_context *c, enum closetype ct)
1118 /* Close off all variables that are in_scope */
1119 struct variable *v, **vp, *v2;
1122 for (vp = &c->in_scope;
1123 v = *vp, v && v->depth > c->scope_depth && v->min_depth > c->scope_depth;
1125 if (v->name->var == v) switch (ct) {
1127 case CloseParallel: /* handle PendingScope */
1131 if (c->scope_stack->child_count == 1)
1132 v->scope = PendingScope;
1133 else if (v->previous &&
1134 v->previous->scope == PendingScope)
1135 v->scope = PendingScope;
1136 else if (v->type == Tlabel)
1137 v->scope = PendingScope;
1138 else if (v->name->var == v)
1139 v->scope = OutScope;
1140 if (ct == CloseElse) {
1141 /* All Pending variables with this name
1142 * are now Conditional */
1144 v2 && v2->scope == PendingScope;
1146 v2->scope = CondScope;
1151 v2 && v2->scope == PendingScope;
1153 if (v2->type != Tlabel)
1154 v2->scope = OutScope;
1156 case OutScope: break;
1159 case CloseSequential:
1160 if (v->type == Tlabel)
1161 v->scope = PendingScope;
1164 v->scope = OutScope;
1167 /* There was no 'else', so we can only become
1168 * conditional if we know the cases were exhaustive,
1169 * and that doesn't mean anything yet.
1170 * So only labels become conditional..
1173 v2 && v2->scope == PendingScope;
1175 if (v2->type == Tlabel) {
1176 v2->scope = CondScope;
1177 v2->min_depth = c->scope_depth;
1179 v2->scope = OutScope;
1182 case OutScope: break;
1186 if (v->scope == OutScope || v->name->var != v)
1195 Executables can be lots of different things. In many cases an
1196 executable is just an operation combined with one or two other
1197 executables. This allows for expressions and lists etc. Other times an
1198 executable is something quite specific like a constant or variable name.
1199 So we define a `struct exec` to be a general executable with a type, and
1200 a `struct binode` which is a subclass of `exec`, forms a node in a
1201 binary tree, and holds an operation. There will be other subclasses,
1202 and to access these we need to be able to `cast` the `exec` into the
1203 various other types. The first field in any `struct exec` is the type
1204 from the `exec_types` enum.
1207 #define cast(structname, pointer) ({ \
1208 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
1209 if (__mptr && *__mptr != X##structname) abort(); \
1210 (struct structname *)( (char *)__mptr);})
1212 #define new(structname) ({ \
1213 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1214 __ptr->type = X##structname; \
1215 __ptr->line = -1; __ptr->column = -1; \
1218 #define new_pos(structname, token) ({ \
1219 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1220 __ptr->type = X##structname; \
1221 __ptr->line = token.line; __ptr->column = token.col; \
1230 enum exec_types type;
1238 struct exec *left, *right;
1241 ###### ast functions
1243 static int __fput_loc(struct exec *loc, FILE *f)
1247 if (loc->line >= 0) {
1248 fprintf(f, "%d:%d: ", loc->line, loc->column);
1251 if (loc->type == Xbinode)
1252 return __fput_loc(cast(binode,loc)->left, f) ||
1253 __fput_loc(cast(binode,loc)->right, f);
1256 static void fput_loc(struct exec *loc, FILE *f)
1258 if (!__fput_loc(loc, f))
1259 fprintf(f, "??:??: "); // NOTEST
1262 Each different type of `exec` node needs a number of functions defined,
1263 a bit like methods. We must be able to free it, print it, analyse it
1264 and execute it. Once we have specific `exec` types we will need to
1265 parse them too. Let's take this a bit more slowly.
1269 The parser generator requires a `free_foo` function for each struct
1270 that stores attributes and they will often be `exec`s and subtypes
1271 there-of. So we need `free_exec` which can handle all the subtypes,
1272 and we need `free_binode`.
1274 ###### ast functions
1276 static void free_binode(struct binode *b)
1281 free_exec(b->right);
1285 ###### core functions
1286 static void free_exec(struct exec *e)
1295 ###### forward decls
1297 static void free_exec(struct exec *e);
1299 ###### free exec cases
1300 case Xbinode: free_binode(cast(binode, e)); break;
1304 Printing an `exec` requires that we know the current indent level for
1305 printing line-oriented components. As will become clear later, we
1306 also want to know what sort of bracketing to use.
1308 ###### ast functions
1310 static void do_indent(int i, char *str)
1317 ###### core functions
1318 static void print_binode(struct binode *b, int indent, int bracket)
1322 ## print binode cases
1326 static void print_exec(struct exec *e, int indent, int bracket)
1332 print_binode(cast(binode, e), indent, bracket); break;
1337 ###### forward decls
1339 static void print_exec(struct exec *e, int indent, int bracket);
1343 As discussed, analysis involves propagating type requirements around the
1344 program and looking for errors.
1346 So `propagate_types` is passed an expected type (being a `struct type`
1347 pointer together with some `val_rules` flags) that the `exec` is
1348 expected to return, and returns the type that it does return, either
1349 of which can be `NULL` signifying "unknown". An `ok` flag is passed
1350 by reference. It is set to `0` when an error is found, and `2` when
1351 any change is made. If it remains unchanged at `1`, then no more
1352 propagation is needed.
1356 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 2<<1};
1360 if (rules & Rnolabel)
1361 fputs(" (labels not permitted)", stderr);
1364 ###### core functions
1366 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1367 struct type *type, int rules);
1368 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1369 struct type *type, int rules)
1376 switch (prog->type) {
1379 struct binode *b = cast(binode, prog);
1381 ## propagate binode cases
1385 ## propagate exec cases
1390 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1391 struct type *type, int rules)
1393 struct type *ret = __propagate_types(prog, c, ok, type, rules);
1402 Interpreting an `exec` doesn't require anything but the `exec`. State
1403 is stored in variables and each variable will be directly linked from
1404 within the `exec` tree. The exception to this is the whole `program`
1405 which needs to look at command line arguments. The `program` will be
1406 interpreted separately.
1408 Each `exec` can return a value combined with a type in `struct lrval`.
1409 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
1410 the location of a value, which can be updated, in `lval`. Others will
1411 set `lval` to NULL indicating that there is a value of appropriate type
1415 ###### core functions
1419 struct value rval, *lval;
1422 static struct lrval _interp_exec(struct exec *e);
1424 static struct value interp_exec(struct exec *e, struct type **typeret)
1426 struct lrval ret = _interp_exec(e);
1428 if (!ret.type) abort();
1430 *typeret = ret.type;
1432 dup_value(ret.type, ret.lval, &ret.rval);
1436 static struct value *linterp_exec(struct exec *e, struct type **typeret)
1438 struct lrval ret = _interp_exec(e);
1441 *typeret = ret.type;
1445 static struct lrval _interp_exec(struct exec *e)
1448 struct value rv = {}, *lrv = NULL;
1449 struct type *rvtype;
1451 rvtype = ret.type = Tnone;
1461 struct binode *b = cast(binode, e);
1462 struct value left, right, *lleft;
1463 struct type *ltype, *rtype;
1464 ltype = rtype = Tnone;
1466 ## interp binode cases
1468 free_value(ltype, &left);
1469 free_value(rtype, &right);
1472 ## interp exec cases
1482 Now that we have the shape of the interpreter in place we can add some
1483 complex types and connected them in to the data structures and the
1484 different phases of parse, analyse, print, interpret.
1486 Thus far we have arrays and structs.
1490 Arrays can be declared by giving a size and a type, as `[size]type' so
1491 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
1492 size can be either a literal number, or a named constant. Some day an
1493 arbitrary expression will be supported.
1495 Arrays cannot be assigned. When pointers are introduced we will also
1496 introduce array slices which can refer to part or all of an array -
1497 the assignment syntax will create a slice. For now, an array can only
1498 ever be referenced by the name it is declared with. It is likely that
1499 a "`copy`" primitive will eventually be define which can be used to
1500 make a copy of an array with controllable recursive depth.
1502 ###### type union fields
1506 struct variable *vsize;
1507 struct type *member;
1510 ###### value functions
1512 static void array_init(struct type *type, struct value *val)
1516 if (type->array.vsize) {
1519 mpz_tdiv_q(q, mpq_numref(type->array.vsize->val->num),
1520 mpq_denref(type->array.vsize->val->num));
1521 type->array.size = mpz_get_si(q);
1524 type->size = type->array.size * type->array.member->size;
1525 type->align = type->array.member->align;
1529 for (i = 0; i < type->array.size; i++) {
1531 v = (void*)val->ptr + i * type->array.member->size;
1532 val_init(type->array.member, v);
1536 static void array_free(struct type *type, struct value *val)
1540 for (i = 0; i < type->array.size; i++) {
1542 v = (void*)val->ptr + i * type->array.member->size;
1543 free_value(type->array.member, v);
1547 static int array_compat(struct type *require, struct type *have)
1549 if (have->compat != require->compat)
1551 /* Both are arrays, so we can look at details */
1552 if (!type_compat(require->array.member, have->array.member, 0))
1554 if (require->array.vsize == NULL && have->array.vsize == NULL)
1555 return require->array.size == have->array.size;
1557 return require->array.vsize == have->array.vsize;
1560 static void array_print_type(struct type *type, FILE *f)
1563 if (type->array.vsize) {
1564 struct binding *b = type->array.vsize->name;
1565 fprintf(f, "%.*s]", b->name.len, b->name.txt);
1567 fprintf(f, "%d]", type->array.size);
1568 type_print(type->array.member, f);
1571 static struct type array_prototype = {
1573 .print_type = array_print_type,
1574 .compat = array_compat,
1578 ###### declare terminals
1583 | [ NUMBER ] Type ${ {
1586 struct text noname = { "", 0 };
1588 $0 = add_type(c, noname, &array_prototype);
1589 $0->array.member = $<4;
1590 $0->array.vsize = NULL;
1591 if (number_parse(num, tail, $2.txt) == 0)
1592 tok_err(c, "error: unrecognised number", &$2);
1594 tok_err(c, "error: unsupported number suffix", &$2);
1596 $0->array.size = mpz_get_ui(mpq_numref(num));
1597 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
1598 tok_err(c, "error: array size must be an integer",
1600 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
1601 tok_err(c, "error: array size is too large",
1607 | [ IDENTIFIER ] Type ${ {
1608 struct variable *v = var_ref(c, $2.txt);
1609 struct text noname = { "", 0 };
1612 tok_err(c, "error: name undeclared", &$2);
1613 else if (!v->constant)
1614 tok_err(c, "error: array size must be a constant", &$2);
1616 $0 = add_type(c, noname, &array_prototype);
1617 $0->array.member = $<4;
1619 $0->array.vsize = v;
1625 ###### variable grammar
1627 | Variable [ Expression ] ${ {
1628 struct binode *b = new(binode);
1635 ###### print binode cases
1637 print_exec(b->left, -1, bracket);
1639 print_exec(b->right, -1, bracket);
1643 ###### propagate binode cases
1645 /* left must be an array, right must be a number,
1646 * result is the member type of the array
1648 propagate_types(b->right, c, ok, Tnum, 0);
1649 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
1650 if (!t || t->compat != array_compat) {
1651 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
1654 if (!type_compat(type, t->array.member, rules)) {
1655 type_err(c, "error: have %1 but need %2", prog,
1656 t->array.member, rules, type);
1658 return t->array.member;
1662 ###### interp binode cases
1667 lleft = linterp_exec(b->left, <ype);
1668 right = interp_exec(b->right, &rtype);
1670 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
1674 rvtype = ltype->array.member;
1675 if (i >= 0 && i < ltype->array.size)
1676 lrv = (void*)lleft + i * rvtype->size;
1678 val_init(ltype->array.member, &rv);
1685 A `struct` is a data-type that contains one or more other data-types.
1686 It differs from an array in that each member can be of a different
1687 type, and they are accessed by name rather than by number. Thus you
1688 cannot choose an element by calculation, you need to know what you
1691 The language makes no promises about how a given structure will be
1692 stored in memory - it is free to rearrange fields to suit whatever
1693 criteria seems important.
1695 Structs are declared separately from program code - they cannot be
1696 declared in-line in a variable declaration like arrays can. A struct
1697 is given a name and this name is used to identify the type - the name
1698 is not prefixed by the word `struct` as it would be in C.
1700 Structs are only treated as the same if they have the same name.
1701 Simply having the same fields in the same order is not enough. This
1702 might change once we can create structure initializers from a list of
1705 Each component datum is identified much like a variable is declared,
1706 with a name, one or two colons, and a type. The type cannot be omitted
1707 as there is no opportunity to deduce the type from usage. An initial
1708 value can be given following an equals sign, so
1710 ##### Example: a struct type
1716 would declare a type called "complex" which has two number fields,
1717 each initialised to zero.
1719 Struct will need to be declared separately from the code that uses
1720 them, so we will need to be able to print out the declaration of a
1721 struct when reprinting the whole program. So a `print_type_decl` type
1722 function will be needed.
1724 ###### type union fields
1736 ###### type functions
1737 void (*print_type_decl)(struct type *type, FILE *f);
1739 ###### value functions
1741 static void structure_init(struct type *type, struct value *val)
1745 for (i = 0; i < type->structure.nfields; i++) {
1747 v = (void*) val->ptr + type->structure.fields[i].offset;
1748 val_init(type->structure.fields[i].type, v);
1752 static void structure_free(struct type *type, struct value *val)
1756 for (i = 0; i < type->structure.nfields; i++) {
1758 v = (void*)val->ptr + type->structure.fields[i].offset;
1759 free_value(type->structure.fields[i].type, v);
1763 static void structure_free_type(struct type *t)
1766 for (i = 0; i < t->structure.nfields; i++)
1767 if (t->structure.fields[i].init) {
1768 free_value(t->structure.fields[i].type,
1769 t->structure.fields[i].init);
1770 free(t->structure.fields[i].init);
1772 free(t->structure.fields);
1775 static struct type structure_prototype = {
1776 .init = structure_init,
1777 .free = structure_free,
1778 .free_type = structure_free_type,
1779 .print_type_decl = structure_print_type,
1793 ###### free exec cases
1795 free_exec(cast(fieldref, e)->left);
1799 ###### declare terminals
1802 ###### variable grammar
1804 | Variable . IDENTIFIER ${ {
1805 struct fieldref *fr = new_pos(fieldref, $2);
1812 ###### print exec cases
1816 struct fieldref *f = cast(fieldref, e);
1817 print_exec(f->left, -1, bracket);
1818 printf(".%.*s", f->name.len, f->name.txt);
1822 ###### ast functions
1823 static int find_struct_index(struct type *type, struct text field)
1826 for (i = 0; i < type->structure.nfields; i++)
1827 if (text_cmp(type->structure.fields[i].name, field) == 0)
1832 ###### propagate exec cases
1836 struct fieldref *f = cast(fieldref, prog);
1837 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
1840 type_err(c, "error: unknown type for field access", f->left,
1842 else if (st->init != structure_init)
1843 type_err(c, "error: field reference attempted on %1, not a struct",
1844 f->left, st, 0, NULL);
1845 else if (f->index == -2) {
1846 f->index = find_struct_index(st, f->name);
1848 type_err(c, "error: cannot find requested field in %1",
1849 f->left, st, 0, NULL);
1851 if (f->index >= 0) {
1852 struct type *ft = st->structure.fields[f->index].type;
1853 if (!type_compat(type, ft, rules))
1854 type_err(c, "error: have %1 but need %2", prog,
1861 ###### interp exec cases
1864 struct fieldref *f = cast(fieldref, e);
1866 struct value *lleft = linterp_exec(f->left, <ype);
1867 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
1868 rvtype = ltype->structure.fields[f->index].type;
1874 struct fieldlist *prev;
1878 ###### ast functions
1879 static void free_fieldlist(struct fieldlist *f)
1883 free_fieldlist(f->prev);
1885 free_value(f->f.type, f->f.init);
1891 ###### top level grammar
1892 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
1894 add_type(c, $2.txt, &structure_prototype);
1896 struct fieldlist *f;
1898 for (f = $3; f; f=f->prev)
1901 t->structure.nfields = cnt;
1902 t->structure.fields = calloc(cnt, sizeof(struct field));
1905 int a = f->f.type->align;
1907 t->structure.fields[cnt] = f->f;
1908 if (t->size & (a-1))
1909 t->size = (t->size | (a-1)) + 1;
1910 t->structure.fields[cnt].offset = t->size;
1911 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
1920 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
1921 | { SimpleFieldList } ${ $0 = $<SFL; }$
1922 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
1923 | SimpleFieldList EOL ${ $0 = $<SFL; }$
1925 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
1926 | FieldLines SimpleFieldList Newlines ${
1931 SimpleFieldList -> Field ${ $0 = $<F; }$
1932 | SimpleFieldList ; Field ${
1936 | SimpleFieldList ; ${
1939 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
1941 Field -> IDENTIFIER : Type = Expression ${ {
1944 $0 = calloc(1, sizeof(struct fieldlist));
1945 $0->f.name = $1.txt;
1950 propagate_types($<5, c, &ok, $3, 0);
1955 struct value vl = interp_exec($5, NULL);
1956 $0->f.init = val_alloc($0->f.type, &vl);
1959 | IDENTIFIER : Type ${
1960 $0 = calloc(1, sizeof(struct fieldlist));
1961 $0->f.name = $1.txt;
1963 $0->f.init = val_alloc($0->f.type, NULL);
1966 ###### forward decls
1967 static void structure_print_type(struct type *t, FILE *f);
1969 ###### value functions
1970 static void structure_print_type(struct type *t, FILE *f)
1974 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
1976 for (i = 0; i < t->structure.nfields; i++) {
1977 struct field *fl = t->structure.fields + i;
1978 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
1979 type_print(fl->type, f);
1980 if (fl->type->print && fl->init) {
1982 if (fl->type == Tstr)
1984 print_value(fl->type, fl->init);
1985 if (fl->type == Tstr)
1992 ###### print type decls
1997 while (target != 0) {
1999 for (t = context.typelist; t ; t=t->next)
2000 if (t->print_type_decl) {
2009 t->print_type_decl(t, stdout);
2015 ## Executables: the elements of code
2017 Each code element needs to be parsed, printed, analysed,
2018 interpreted, and freed. There are several, so let's just start with
2019 the easy ones and work our way up.
2023 We have already met values as separate objects. When manifest
2024 constants appear in the program text, that must result in an executable
2025 which has a constant value. So the `val` structure embeds a value in
2038 ###### ast functions
2039 struct val *new_val(struct type *T, struct token tk)
2041 struct val *v = new_pos(val, tk);
2052 $0 = new_val(Tbool, $1);
2056 $0 = new_val(Tbool, $1);
2060 $0 = new_val(Tnum, $1);
2063 if (number_parse($0->val.num, tail, $1.txt) == 0)
2064 mpq_init($0->val.num);
2066 tok_err(c, "error: unsupported number suffix",
2071 $0 = new_val(Tstr, $1);
2074 string_parse(&$1, '\\', &$0->val.str, tail);
2076 tok_err(c, "error: unsupported string suffix",
2081 $0 = new_val(Tstr, $1);
2084 string_parse(&$1, '\\', &$0->val.str, tail);
2086 tok_err(c, "error: unsupported string suffix",
2091 ###### print exec cases
2094 struct val *v = cast(val, e);
2095 if (v->vtype == Tstr)
2097 print_value(v->vtype, &v->val);
2098 if (v->vtype == Tstr)
2103 ###### propagate exec cases
2106 struct val *val = cast(val, prog);
2107 if (!type_compat(type, val->vtype, rules))
2108 type_err(c, "error: expected %1%r found %2",
2109 prog, type, rules, val->vtype);
2113 ###### interp exec cases
2115 rvtype = cast(val, e)->vtype;
2116 dup_value(rvtype, &cast(val, e)->val, &rv);
2119 ###### ast functions
2120 static void free_val(struct val *v)
2123 free_value(v->vtype, &v->val);
2127 ###### free exec cases
2128 case Xval: free_val(cast(val, e)); break;
2130 ###### ast functions
2131 // Move all nodes from 'b' to 'rv', reversing their order.
2132 // In 'b' 'left' is a list, and 'right' is the last node.
2133 // In 'rv', left' is the first node and 'right' is a list.
2134 static struct binode *reorder_bilist(struct binode *b)
2136 struct binode *rv = NULL;
2139 struct exec *t = b->right;
2143 b = cast(binode, b->left);
2153 Just as we used a `val` to wrap a value into an `exec`, we similarly
2154 need a `var` to wrap a `variable` into an exec. While each `val`
2155 contained a copy of the value, each `var` holds a link to the variable
2156 because it really is the same variable no matter where it appears.
2157 When a variable is used, we need to remember to follow the `->merged`
2158 link to find the primary instance.
2166 struct variable *var;
2174 VariableDecl -> IDENTIFIER : ${ {
2175 struct variable *v = var_decl(c, $1.txt);
2176 $0 = new_pos(var, $1);
2181 v = var_ref(c, $1.txt);
2183 type_err(c, "error: variable '%v' redeclared",
2185 type_err(c, "info: this is where '%v' was first declared",
2186 v->where_decl, NULL, 0, NULL);
2189 | IDENTIFIER :: ${ {
2190 struct variable *v = var_decl(c, $1.txt);
2191 $0 = new_pos(var, $1);
2197 v = var_ref(c, $1.txt);
2199 type_err(c, "error: variable '%v' redeclared",
2201 type_err(c, "info: this is where '%v' was first declared",
2202 v->where_decl, NULL, 0, NULL);
2205 | IDENTIFIER : Type ${ {
2206 struct variable *v = var_decl(c, $1.txt);
2207 $0 = new_pos(var, $1);
2215 v = var_ref(c, $1.txt);
2217 type_err(c, "error: variable '%v' redeclared",
2219 type_err(c, "info: this is where '%v' was first declared",
2220 v->where_decl, NULL, 0, NULL);
2223 | IDENTIFIER :: Type ${ {
2224 struct variable *v = var_decl(c, $1.txt);
2225 $0 = new_pos(var, $1);
2234 v = var_ref(c, $1.txt);
2236 type_err(c, "error: variable '%v' redeclared",
2238 type_err(c, "info: this is where '%v' was first declared",
2239 v->where_decl, NULL, 0, NULL);
2244 Variable -> IDENTIFIER ${ {
2245 struct variable *v = var_ref(c, $1.txt);
2246 $0 = new_pos(var, $1);
2248 /* This might be a label - allocate a var just in case */
2249 v = var_decl(c, $1.txt);
2257 cast(var, $0)->var = v;
2262 Type -> IDENTIFIER ${
2263 $0 = find_type(c, $1.txt);
2266 "error: undefined type", &$1);
2273 ###### print exec cases
2276 struct var *v = cast(var, e);
2278 struct binding *b = v->var->name;
2279 printf("%.*s", b->name.len, b->name.txt);
2286 if (loc->type == Xvar) {
2287 struct var *v = cast(var, loc);
2289 struct binding *b = v->var->name;
2290 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2292 fputs("???", stderr); // NOTEST
2294 fputs("NOTVAR", stderr); // NOTEST
2297 ###### propagate exec cases
2301 struct var *var = cast(var, prog);
2302 struct variable *v = var->var;
2304 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2305 return Tnone; // NOTEST
2309 if (v->constant && (rules & Rnoconstant)) {
2310 type_err(c, "error: Cannot assign to a constant: %v",
2311 prog, NULL, 0, NULL);
2312 type_err(c, "info: name was defined as a constant here",
2313 v->where_decl, NULL, 0, NULL);
2316 if (v->type == Tnone && v->where_decl == prog)
2317 type_err(c, "error: variable used but not declared: %v",
2318 prog, NULL, 0, NULL);
2319 if (v->type == NULL) {
2320 if (type && *ok != 0) {
2323 v->where_set = prog;
2328 if (!type_compat(type, v->type, rules)) {
2329 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2330 type, rules, v->type);
2331 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2332 v->type, rules, NULL);
2339 ###### interp exec cases
2342 struct var *var = cast(var, e);
2343 struct variable *v = var->var;
2352 ###### ast functions
2354 static void free_var(struct var *v)
2359 ###### free exec cases
2360 case Xvar: free_var(cast(var, e)); break;
2362 ### Expressions: Conditional
2364 Our first user of the `binode` will be conditional expressions, which
2365 is a bit odd as they actually have three components. That will be
2366 handled by having 2 binodes for each expression. The conditional
2367 expression is the lowest precedence operator which is why we define it
2368 first - to start the precedence list.
2370 Conditional expressions are of the form "value `if` condition `else`
2371 other_value". They associate to the right, so everything to the right
2372 of `else` is part of an else value, while only a higher-precedence to
2373 the left of `if` is the if values. Between `if` and `else` there is no
2374 room for ambiguity, so a full conditional expression is allowed in
2386 Expression -> Expression if Expression else Expression $$ifelse ${ {
2387 struct binode *b1 = new(binode);
2388 struct binode *b2 = new(binode);
2397 ## expression grammar
2399 ###### print binode cases
2402 b2 = cast(binode, b->right);
2403 if (bracket) printf("(");
2404 print_exec(b2->left, -1, bracket);
2406 print_exec(b->left, -1, bracket);
2408 print_exec(b2->right, -1, bracket);
2409 if (bracket) printf(")");
2412 ###### propagate binode cases
2415 /* cond must be Tbool, others must match */
2416 struct binode *b2 = cast(binode, b->right);
2419 propagate_types(b->left, c, ok, Tbool, 0);
2420 t = propagate_types(b2->left, c, ok, type, Rnolabel);
2421 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
2425 ###### interp binode cases
2428 struct binode *b2 = cast(binode, b->right);
2429 left = interp_exec(b->left, <ype);
2431 rv = interp_exec(b2->left, &rvtype);
2433 rv = interp_exec(b2->right, &rvtype);
2437 ### Expressions: Boolean
2439 The next class of expressions to use the `binode` will be Boolean
2440 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
2441 have same corresponding precendence. The difference is that they don't
2442 evaluate the second expression if not necessary.
2451 ###### expr precedence
2456 ###### expression grammar
2457 | Expression or Expression ${ {
2458 struct binode *b = new(binode);
2464 | Expression or else Expression ${ {
2465 struct binode *b = new(binode);
2472 | Expression and Expression ${ {
2473 struct binode *b = new(binode);
2479 | Expression and then Expression ${ {
2480 struct binode *b = new(binode);
2487 | not Expression ${ {
2488 struct binode *b = new(binode);
2494 ###### print binode cases
2496 if (bracket) printf("(");
2497 print_exec(b->left, -1, bracket);
2499 print_exec(b->right, -1, bracket);
2500 if (bracket) printf(")");
2503 if (bracket) printf("(");
2504 print_exec(b->left, -1, bracket);
2505 printf(" and then ");
2506 print_exec(b->right, -1, bracket);
2507 if (bracket) printf(")");
2510 if (bracket) printf("(");
2511 print_exec(b->left, -1, bracket);
2513 print_exec(b->right, -1, bracket);
2514 if (bracket) printf(")");
2517 if (bracket) printf("(");
2518 print_exec(b->left, -1, bracket);
2519 printf(" or else ");
2520 print_exec(b->right, -1, bracket);
2521 if (bracket) printf(")");
2524 if (bracket) printf("(");
2526 print_exec(b->right, -1, bracket);
2527 if (bracket) printf(")");
2530 ###### propagate binode cases
2536 /* both must be Tbool, result is Tbool */
2537 propagate_types(b->left, c, ok, Tbool, 0);
2538 propagate_types(b->right, c, ok, Tbool, 0);
2539 if (type && type != Tbool)
2540 type_err(c, "error: %1 operation found where %2 expected", prog,
2544 ###### interp binode cases
2546 rv = interp_exec(b->left, &rvtype);
2547 right = interp_exec(b->right, &rtype);
2548 rv.bool = rv.bool && right.bool;
2551 rv = interp_exec(b->left, &rvtype);
2553 rv = interp_exec(b->right, NULL);
2556 rv = interp_exec(b->left, &rvtype);
2557 right = interp_exec(b->right, &rtype);
2558 rv.bool = rv.bool || right.bool;
2561 rv = interp_exec(b->left, &rvtype);
2563 rv = interp_exec(b->right, NULL);
2566 rv = interp_exec(b->right, &rvtype);
2570 ### Expressions: Comparison
2572 Of slightly higher precedence that Boolean expressions are Comparisons.
2573 A comparison takes arguments of any comparable type, but the two types
2576 To simplify the parsing we introduce an `eop` which can record an
2577 expression operator, and the `CMPop` non-terminal will match one of them.
2584 ###### ast functions
2585 static void free_eop(struct eop *e)
2599 ###### expr precedence
2600 $LEFT < > <= >= == != CMPop
2602 ###### expression grammar
2603 | Expression CMPop Expression ${ {
2604 struct binode *b = new(binode);
2614 CMPop -> < ${ $0.op = Less; }$
2615 | > ${ $0.op = Gtr; }$
2616 | <= ${ $0.op = LessEq; }$
2617 | >= ${ $0.op = GtrEq; }$
2618 | == ${ $0.op = Eql; }$
2619 | != ${ $0.op = NEql; }$
2621 ###### print binode cases
2629 if (bracket) printf("(");
2630 print_exec(b->left, -1, bracket);
2632 case Less: printf(" < "); break;
2633 case LessEq: printf(" <= "); break;
2634 case Gtr: printf(" > "); break;
2635 case GtrEq: printf(" >= "); break;
2636 case Eql: printf(" == "); break;
2637 case NEql: printf(" != "); break;
2638 default: abort(); // NOTEST
2640 print_exec(b->right, -1, bracket);
2641 if (bracket) printf(")");
2644 ###### propagate binode cases
2651 /* Both must match but not be labels, result is Tbool */
2652 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
2654 propagate_types(b->right, c, ok, t, 0);
2656 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
2658 t = propagate_types(b->left, c, ok, t, 0);
2660 if (!type_compat(type, Tbool, 0))
2661 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
2662 Tbool, rules, type);
2665 ###### interp binode cases
2674 left = interp_exec(b->left, <ype);
2675 right = interp_exec(b->right, &rtype);
2676 cmp = value_cmp(ltype, rtype, &left, &right);
2679 case Less: rv.bool = cmp < 0; break;
2680 case LessEq: rv.bool = cmp <= 0; break;
2681 case Gtr: rv.bool = cmp > 0; break;
2682 case GtrEq: rv.bool = cmp >= 0; break;
2683 case Eql: rv.bool = cmp == 0; break;
2684 case NEql: rv.bool = cmp != 0; break;
2685 default: rv.bool = 0; break; // NOTEST
2690 ### Expressions: The rest
2692 The remaining expressions with the highest precedence are arithmetic,
2693 string concatenation, and string conversion. String concatenation
2694 (`++`) has the same precedence as multiplication and division, but lower
2697 String conversion is a temporary feature until I get a better type
2698 system. `$` is a prefix operator which expects a string and returns
2701 `+` and `-` are both infix and prefix operations (where they are
2702 absolute value and negation). These have different operator names.
2704 We also have a 'Bracket' operator which records where parentheses were
2705 found. This makes it easy to reproduce these when printing. Possibly I
2706 should only insert brackets were needed for precedence.
2716 ###### expr precedence
2722 ###### expression grammar
2723 | Expression Eop Expression ${ {
2724 struct binode *b = new(binode);
2731 | Expression Top Expression ${ {
2732 struct binode *b = new(binode);
2739 | ( Expression ) ${ {
2740 struct binode *b = new_pos(binode, $1);
2745 | Uop Expression ${ {
2746 struct binode *b = new(binode);
2751 | Value ${ $0 = $<1; }$
2752 | Variable ${ $0 = $<1; }$
2755 Eop -> + ${ $0.op = Plus; }$
2756 | - ${ $0.op = Minus; }$
2758 Uop -> + ${ $0.op = Absolute; }$
2759 | - ${ $0.op = Negate; }$
2760 | $ ${ $0.op = StringConv; }$
2762 Top -> * ${ $0.op = Times; }$
2763 | / ${ $0.op = Divide; }$
2764 | % ${ $0.op = Rem; }$
2765 | ++ ${ $0.op = Concat; }$
2767 ###### print binode cases
2774 if (bracket) printf("(");
2775 print_exec(b->left, indent, bracket);
2777 case Plus: fputs(" + ", stdout); break;
2778 case Minus: fputs(" - ", stdout); break;
2779 case Times: fputs(" * ", stdout); break;
2780 case Divide: fputs(" / ", stdout); break;
2781 case Rem: fputs(" % ", stdout); break;
2782 case Concat: fputs(" ++ ", stdout); break;
2783 default: abort(); // NOTEST
2785 print_exec(b->right, indent, bracket);
2786 if (bracket) printf(")");
2791 if (bracket) printf("(");
2793 case Absolute: fputs("+", stdout); break;
2794 case Negate: fputs("-", stdout); break;
2795 case StringConv: fputs("$", stdout); break;
2796 default: abort(); // NOTEST
2798 print_exec(b->right, indent, bracket);
2799 if (bracket) printf(")");
2803 print_exec(b->right, indent, bracket);
2807 ###### propagate binode cases
2813 /* both must be numbers, result is Tnum */
2816 /* as propagate_types ignores a NULL,
2817 * unary ops fit here too */
2818 propagate_types(b->left, c, ok, Tnum, 0);
2819 propagate_types(b->right, c, ok, Tnum, 0);
2820 if (!type_compat(type, Tnum, 0))
2821 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
2826 /* both must be Tstr, result is Tstr */
2827 propagate_types(b->left, c, ok, Tstr, 0);
2828 propagate_types(b->right, c, ok, Tstr, 0);
2829 if (!type_compat(type, Tstr, 0))
2830 type_err(c, "error: Concat returns %1 but %2 expected", prog,
2835 /* op must be string, result is number */
2836 propagate_types(b->left, c, ok, Tstr, 0);
2837 if (!type_compat(type, Tnum, 0))
2839 "error: Can only convert string to number, not %1",
2840 prog, type, 0, NULL);
2844 return propagate_types(b->right, c, ok, type, 0);
2846 ###### interp binode cases
2849 rv = interp_exec(b->left, &rvtype);
2850 right = interp_exec(b->right, &rtype);
2851 mpq_add(rv.num, rv.num, right.num);
2854 rv = interp_exec(b->left, &rvtype);
2855 right = interp_exec(b->right, &rtype);
2856 mpq_sub(rv.num, rv.num, right.num);
2859 rv = interp_exec(b->left, &rvtype);
2860 right = interp_exec(b->right, &rtype);
2861 mpq_mul(rv.num, rv.num, right.num);
2864 rv = interp_exec(b->left, &rvtype);
2865 right = interp_exec(b->right, &rtype);
2866 mpq_div(rv.num, rv.num, right.num);
2871 left = interp_exec(b->left, <ype);
2872 right = interp_exec(b->right, &rtype);
2873 mpz_init(l); mpz_init(r); mpz_init(rem);
2874 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
2875 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
2876 mpz_tdiv_r(rem, l, r);
2877 val_init(Tnum, &rv);
2878 mpq_set_z(rv.num, rem);
2879 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
2884 rv = interp_exec(b->right, &rvtype);
2885 mpq_neg(rv.num, rv.num);
2888 rv = interp_exec(b->right, &rvtype);
2889 mpq_abs(rv.num, rv.num);
2892 rv = interp_exec(b->right, &rvtype);
2895 left = interp_exec(b->left, <ype);
2896 right = interp_exec(b->right, &rtype);
2898 rv.str = text_join(left.str, right.str);
2901 right = interp_exec(b->right, &rvtype);
2905 struct text tx = right.str;
2908 if (tx.txt[0] == '-') {
2913 if (number_parse(rv.num, tail, tx) == 0)
2916 mpq_neg(rv.num, rv.num);
2918 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt);
2922 ###### value functions
2924 static struct text text_join(struct text a, struct text b)
2927 rv.len = a.len + b.len;
2928 rv.txt = malloc(rv.len);
2929 memcpy(rv.txt, a.txt, a.len);
2930 memcpy(rv.txt+a.len, b.txt, b.len);
2934 ### Blocks, Statements, and Statement lists.
2936 Now that we have expressions out of the way we need to turn to
2937 statements. There are simple statements and more complex statements.
2938 Simple statements do not contain (syntactic) newlines, complex statements do.
2940 Statements often come in sequences and we have corresponding simple
2941 statement lists and complex statement lists.
2942 The former comprise only simple statements separated by semicolons.
2943 The later comprise complex statements and simple statement lists. They are
2944 separated by newlines. Thus the semicolon is only used to separate
2945 simple statements on the one line. This may be overly restrictive,
2946 but I'm not sure I ever want a complex statement to share a line with
2949 Note that a simple statement list can still use multiple lines if
2950 subsequent lines are indented, so
2952 ###### Example: wrapped simple statement list
2957 is a single simple statement list. This might allow room for
2958 confusion, so I'm not set on it yet.
2960 A simple statement list needs no extra syntax. A complex statement
2961 list has two syntactic forms. It can be enclosed in braces (much like
2962 C blocks), or it can be introduced by an indent and continue until an
2963 unindented newline (much like Python blocks). With this extra syntax
2964 it is referred to as a block.
2966 Note that a block does not have to include any newlines if it only
2967 contains simple statements. So both of:
2969 if condition: a=b; d=f
2971 if condition { a=b; print f }
2975 In either case the list is constructed from a `binode` list with
2976 `Block` as the operator. When parsing the list it is most convenient
2977 to append to the end, so a list is a list and a statement. When using
2978 the list it is more convenient to consider a list to be a statement
2979 and a list. So we need a function to re-order a list.
2980 `reorder_bilist` serves this purpose.
2982 The only stand-alone statement we introduce at this stage is `pass`
2983 which does nothing and is represented as a `NULL` pointer in a `Block`
2984 list. Other stand-alone statements will follow once the infrastructure
2995 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
2996 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
2997 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
2998 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
2999 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3001 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3002 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3003 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3004 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3005 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3007 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3008 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3009 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3011 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3012 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3013 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3014 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3015 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3017 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3019 ComplexStatements -> ComplexStatements ComplexStatement ${
3029 | ComplexStatement ${
3041 ComplexStatement -> SimpleStatements Newlines ${
3042 $0 = reorder_bilist($<SS);
3044 | SimpleStatements ; Newlines ${
3045 $0 = reorder_bilist($<SS);
3047 ## ComplexStatement Grammar
3050 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3056 | SimpleStatement ${
3064 SimpleStatement -> pass ${ $0 = NULL; }$
3065 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3066 ## SimpleStatement Grammar
3068 ###### print binode cases
3072 if (b->left == NULL)
3075 print_exec(b->left, indent, bracket);
3078 print_exec(b->right, indent, bracket);
3081 // block, one per line
3082 if (b->left == NULL)
3083 do_indent(indent, "pass\n");
3085 print_exec(b->left, indent, bracket);
3087 print_exec(b->right, indent, bracket);
3091 ###### propagate binode cases
3094 /* If any statement returns something other than Tnone
3095 * or Tbool then all such must return same type.
3096 * As each statement may be Tnone or something else,
3097 * we must always pass NULL (unknown) down, otherwise an incorrect
3098 * error might occur. We never return Tnone unless it is
3103 for (e = b; e; e = cast(binode, e->right)) {
3104 t = propagate_types(e->left, c, ok, NULL, rules);
3105 if ((rules & Rboolok) && t == Tbool)
3107 if (t && t != Tnone && t != Tbool) {
3111 type_err(c, "error: expected %1%r, found %2",
3112 e->left, type, rules, t);
3118 ###### interp binode cases
3120 while (rvtype == Tnone &&
3123 rv = interp_exec(b->left, &rvtype);
3124 b = cast(binode, b->right);
3128 ### The Print statement
3130 `print` is a simple statement that takes a comma-separated list of
3131 expressions and prints the values separated by spaces and terminated
3132 by a newline. No control of formatting is possible.
3134 `print` faces the same list-ordering issue as blocks, and uses the
3140 ##### expr precedence
3143 ###### SimpleStatement Grammar
3145 | print ExpressionList ${
3146 $0 = reorder_bilist($<2);
3148 | print ExpressionList , ${
3153 $0 = reorder_bilist($0);
3164 ExpressionList -> ExpressionList , Expression ${
3177 ###### print binode cases
3180 do_indent(indent, "print");
3184 print_exec(b->left, -1, bracket);
3188 b = cast(binode, b->right);
3194 ###### propagate binode cases
3197 /* don't care but all must be consistent */
3198 propagate_types(b->left, c, ok, NULL, Rnolabel);
3199 propagate_types(b->right, c, ok, NULL, Rnolabel);
3202 ###### interp binode cases
3208 for ( ; b; b = cast(binode, b->right))
3212 left = interp_exec(b->left, <ype);
3213 print_value(ltype, &left);
3214 free_value(ltype, &left);
3225 ###### Assignment statement
3227 An assignment will assign a value to a variable, providing it hasn't
3228 been declared as a constant. The analysis phase ensures that the type
3229 will be correct so the interpreter just needs to perform the
3230 calculation. There is a form of assignment which declares a new
3231 variable as well as assigning a value. If a name is assigned before
3232 it is declared, and error will be raised as the name is created as
3233 `Tlabel` and it is illegal to assign to such names.
3239 ###### declare terminals
3242 ###### SimpleStatement Grammar
3243 | Variable = Expression ${
3249 | VariableDecl = Expression ${
3257 if ($1->var->where_set == NULL) {
3259 "Variable declared with no type or value: %v",
3269 ###### print binode cases
3272 do_indent(indent, "");
3273 print_exec(b->left, indent, bracket);
3275 print_exec(b->right, indent, bracket);
3282 struct variable *v = cast(var, b->left)->var;
3283 do_indent(indent, "");
3284 print_exec(b->left, indent, bracket);
3285 if (cast(var, b->left)->var->constant) {
3286 if (v->where_decl == v->where_set) {
3288 type_print(v->type, stdout);
3293 if (v->where_decl == v->where_set) {
3295 type_print(v->type, stdout);
3302 print_exec(b->right, indent, bracket);
3309 ###### propagate binode cases
3313 /* Both must match and not be labels,
3314 * Type must support 'dup',
3315 * For Assign, left must not be constant.
3318 t = propagate_types(b->left, c, ok, NULL,
3319 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
3324 if (propagate_types(b->right, c, ok, t, 0) != t)
3325 if (b->left->type == Xvar)
3326 type_err(c, "info: variable '%v' was set as %1 here.",
3327 cast(var, b->left)->var->where_set, t, rules, NULL);
3329 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
3331 propagate_types(b->left, c, ok, t,
3332 (b->op == Assign ? Rnoconstant : 0));
3334 if (t && t->dup == NULL)
3335 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
3340 ###### interp binode cases
3343 lleft = linterp_exec(b->left, <ype);
3344 right = interp_exec(b->right, &rtype);
3346 free_value(ltype, lleft);
3347 dup_value(ltype, &right, lleft);
3354 struct variable *v = cast(var, b->left)->var;
3358 right = interp_exec(b->right, &rtype);
3359 free_value(v->type, v->val);
3361 v->val = val_alloc(v->type, &right);
3364 free_value(v->type, v->val);
3365 v->val = val_alloc(v->type, NULL);
3370 ### The `use` statement
3372 The `use` statement is the last "simple" statement. It is needed when
3373 the condition in a conditional statement is a block. `use` works much
3374 like `return` in C, but only completes the `condition`, not the whole
3380 ###### expr precedence
3383 ###### SimpleStatement Grammar
3385 $0 = new_pos(binode, $1);
3388 if ($0->right->type == Xvar) {
3389 struct var *v = cast(var, $0->right);
3390 if (v->var->type == Tnone) {
3391 /* Convert this to a label */
3392 v->var->type = Tlabel;
3393 v->var->val = val_alloc(Tlabel, NULL);
3394 v->var->val->label = v->var->val;
3399 ###### print binode cases
3402 do_indent(indent, "use ");
3403 print_exec(b->right, -1, bracket);
3408 ###### propagate binode cases
3411 /* result matches value */
3412 return propagate_types(b->right, c, ok, type, 0);
3414 ###### interp binode cases
3417 rv = interp_exec(b->right, &rvtype);
3420 ### The Conditional Statement
3422 This is the biggy and currently the only complex statement. This
3423 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
3424 It is comprised of a number of parts, all of which are optional though
3425 set combinations apply. Each part is (usually) a key word (`then` is
3426 sometimes optional) followed by either an expression or a code block,
3427 except the `casepart` which is a "key word and an expression" followed
3428 by a code block. The code-block option is valid for all parts and,
3429 where an expression is also allowed, the code block can use the `use`
3430 statement to report a value. If the code block does not report a value
3431 the effect is similar to reporting `True`.
3433 The `else` and `case` parts, as well as `then` when combined with
3434 `if`, can contain a `use` statement which will apply to some
3435 containing conditional statement. `for` parts, `do` parts and `then`
3436 parts used with `for` can never contain a `use`, except in some
3437 subordinate conditional statement.
3439 If there is a `forpart`, it is executed first, only once.
3440 If there is a `dopart`, then it is executed repeatedly providing
3441 always that the `condpart` or `cond`, if present, does not return a non-True
3442 value. `condpart` can fail to return any value if it simply executes
3443 to completion. This is treated the same as returning `True`.
3445 If there is a `thenpart` it will be executed whenever the `condpart`
3446 or `cond` returns True (or does not return any value), but this will happen
3447 *after* `dopart` (when present).
3449 If `elsepart` is present it will be executed at most once when the
3450 condition returns `False` or some value that isn't `True` and isn't
3451 matched by any `casepart`. If there are any `casepart`s, they will be
3452 executed when the condition returns a matching value.
3454 The particular sorts of values allowed in case parts has not yet been
3455 determined in the language design, so nothing is prohibited.
3457 The various blocks in this complex statement potentially provide scope
3458 for variables as described earlier. Each such block must include the
3459 "OpenScope" nonterminal before parsing the block, and must call
3460 `var_block_close()` when closing the block.
3462 The code following "`if`", "`switch`" and "`for`" does not get its own
3463 scope, but is in a scope covering the whole statement, so names
3464 declared there cannot be redeclared elsewhere. Similarly the
3465 condition following "`while`" is in a scope the covers the body
3466 ("`do`" part) of the loop, and which does not allow conditional scope
3467 extension. Code following "`then`" (both looping and non-looping),
3468 "`else`" and "`case`" each get their own local scope.
3470 The type requirements on the code block in a `whilepart` are quite
3471 unusal. It is allowed to return a value of some identifiable type, in
3472 which case the loop aborts and an appropriate `casepart` is run, or it
3473 can return a Boolean, in which case the loop either continues to the
3474 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
3475 This is different both from the `ifpart` code block which is expected to
3476 return a Boolean, or the `switchpart` code block which is expected to
3477 return the same type as the casepart values. The correct analysis of
3478 the type of the `whilepart` code block is the reason for the
3479 `Rboolok` flag which is passed to `propagate_types()`.
3481 The `cond_statement` cannot fit into a `binode` so a new `exec` is
3490 struct exec *action;
3491 struct casepart *next;
3493 struct cond_statement {
3495 struct exec *forpart, *condpart, *dopart, *thenpart, *elsepart;
3496 struct casepart *casepart;
3499 ###### ast functions
3501 static void free_casepart(struct casepart *cp)
3505 free_exec(cp->value);
3506 free_exec(cp->action);
3513 static void free_cond_statement(struct cond_statement *s)
3517 free_exec(s->forpart);
3518 free_exec(s->condpart);
3519 free_exec(s->dopart);
3520 free_exec(s->thenpart);
3521 free_exec(s->elsepart);
3522 free_casepart(s->casepart);
3526 ###### free exec cases
3527 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
3529 ###### ComplexStatement Grammar
3530 | CondStatement ${ $0 = $<1; }$
3532 ###### expr precedence
3533 $TERM for then while do
3540 // A CondStatement must end with EOL, as does CondSuffix and
3542 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
3543 // may or may not end with EOL
3544 // WhilePart and IfPart include an appropriate Suffix
3547 // Both ForPart and Whilepart open scopes, and CondSuffix only
3548 // closes one - so in the first branch here we have another to close.
3549 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
3552 $0->thenpart = $<TP;
3553 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3554 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3555 var_block_close(c, CloseSequential);
3557 | ForPart OptNL WhilePart CondSuffix ${
3560 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3561 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3562 var_block_close(c, CloseSequential);
3564 | WhilePart CondSuffix ${
3566 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3567 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3569 | SwitchPart OptNL CasePart CondSuffix ${
3571 $0->condpart = $<SP;
3572 $CP->next = $0->casepart;
3573 $0->casepart = $<CP;
3575 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
3577 $0->condpart = $<SP;
3578 $CP->next = $0->casepart;
3579 $0->casepart = $<CP;
3581 | IfPart IfSuffix ${
3583 $0->condpart = $IP.condpart; $IP.condpart = NULL;
3584 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
3585 // This is where we close an "if" statement
3586 var_block_close(c, CloseSequential);
3589 CondSuffix -> IfSuffix ${
3591 // This is where we close scope of the whole
3592 // "for" or "while" statement
3593 var_block_close(c, CloseSequential);
3595 | Newlines CasePart CondSuffix ${
3597 $CP->next = $0->casepart;
3598 $0->casepart = $<CP;
3600 | CasePart CondSuffix ${
3602 $CP->next = $0->casepart;
3603 $0->casepart = $<CP;
3606 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
3607 | Newlines ElsePart ${ $0 = $<EP; }$
3608 | ElsePart ${$0 = $<EP; }$
3610 ElsePart -> else OpenBlock Newlines ${
3611 $0 = new(cond_statement);
3612 $0->elsepart = $<OB;
3613 var_block_close(c, CloseElse);
3615 | else OpenScope CondStatement ${
3616 $0 = new(cond_statement);
3617 $0->elsepart = $<CS;
3618 var_block_close(c, CloseElse);
3622 CasePart -> case Expression OpenScope ColonBlock ${
3623 $0 = calloc(1,sizeof(struct casepart));
3626 var_block_close(c, CloseParallel);
3630 // These scopes are closed in CondSuffix
3631 ForPart -> for OpenBlock ${
3635 ThenPart -> then OpenBlock ${
3637 var_block_close(c, CloseSequential);
3641 // This scope is closed in CondSuffix
3642 WhilePart -> while UseBlock OptNL do Block ${
3646 | while OpenScope Expression ColonBlock ${
3647 $0.condpart = $<Exp;
3651 IfPart -> if UseBlock OptNL then OpenBlock ClosePara ${
3655 | if OpenScope Expression OpenScope ColonBlock ClosePara ${
3659 | if OpenScope Expression OpenScope OptNL then Block ClosePara ${
3665 // This scope is closed in CondSuffix
3666 SwitchPart -> switch OpenScope Expression ${
3669 | switch UseBlock ${
3673 ###### print exec cases
3675 case Xcond_statement:
3677 struct cond_statement *cs = cast(cond_statement, e);
3678 struct casepart *cp;
3680 do_indent(indent, "for");
3681 if (bracket) printf(" {\n"); else printf("\n");
3682 print_exec(cs->forpart, indent+1, bracket);
3685 do_indent(indent, "} then {\n");
3687 do_indent(indent, "then\n");
3688 print_exec(cs->thenpart, indent+1, bracket);
3690 if (bracket) do_indent(indent, "}\n");
3694 if (cs->condpart && cs->condpart->type == Xbinode &&
3695 cast(binode, cs->condpart)->op == Block) {
3697 do_indent(indent, "while {\n");
3699 do_indent(indent, "while\n");
3700 print_exec(cs->condpart, indent+1, bracket);
3702 do_indent(indent, "} do {\n");
3704 do_indent(indent, "do\n");
3705 print_exec(cs->dopart, indent+1, bracket);
3707 do_indent(indent, "}\n");
3709 do_indent(indent, "while ");
3710 print_exec(cs->condpart, 0, bracket);
3715 print_exec(cs->dopart, indent+1, bracket);
3717 do_indent(indent, "}\n");
3722 do_indent(indent, "switch");
3724 do_indent(indent, "if");
3725 if (cs->condpart && cs->condpart->type == Xbinode &&
3726 cast(binode, cs->condpart)->op == Block) {
3731 print_exec(cs->condpart, indent+1, bracket);
3733 do_indent(indent, "}\n");
3735 do_indent(indent, "then:\n");
3736 print_exec(cs->thenpart, indent+1, bracket);
3740 print_exec(cs->condpart, 0, bracket);
3746 print_exec(cs->thenpart, indent+1, bracket);
3748 do_indent(indent, "}\n");
3753 for (cp = cs->casepart; cp; cp = cp->next) {
3754 do_indent(indent, "case ");
3755 print_exec(cp->value, -1, 0);
3760 print_exec(cp->action, indent+1, bracket);
3762 do_indent(indent, "}\n");
3765 do_indent(indent, "else");
3770 print_exec(cs->elsepart, indent+1, bracket);
3772 do_indent(indent, "}\n");
3777 ###### propagate exec cases
3778 case Xcond_statement:
3780 // forpart and dopart must return Tnone
3781 // thenpart must return Tnone if there is a dopart,
3782 // otherwise it is like elsepart.
3784 // be bool if there is no casepart
3785 // match casepart->values if there is a switchpart
3786 // either be bool or match casepart->value if there
3788 // elsepart and casepart->action must match the return type
3789 // expected of this statement.
3790 struct cond_statement *cs = cast(cond_statement, prog);
3791 struct casepart *cp;
3793 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
3794 if (!type_compat(Tnone, t, 0))
3796 t = propagate_types(cs->dopart, c, ok, Tnone, 0);
3797 if (!type_compat(Tnone, t, 0))
3800 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
3801 if (!type_compat(Tnone, t, 0))
3804 if (cs->casepart == NULL)
3805 propagate_types(cs->condpart, c, ok, Tbool, 0);
3807 /* Condpart must match case values, with bool permitted */
3809 for (cp = cs->casepart;
3810 cp && !t; cp = cp->next)
3811 t = propagate_types(cp->value, c, ok, NULL, 0);
3812 if (!t && cs->condpart)
3813 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok);
3814 // Now we have a type (I hope) push it down
3816 for (cp = cs->casepart; cp; cp = cp->next)
3817 propagate_types(cp->value, c, ok, t, 0);
3818 propagate_types(cs->condpart, c, ok, t, Rboolok);
3821 // (if)then, else, and case parts must return expected type.
3822 if (!cs->dopart && !type)
3823 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
3825 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
3826 for (cp = cs->casepart;
3829 type = propagate_types(cp->action, c, ok, NULL, rules);
3832 propagate_types(cs->thenpart, c, ok, type, rules);
3833 propagate_types(cs->elsepart, c, ok, type, rules);
3834 for (cp = cs->casepart; cp ; cp = cp->next)
3835 propagate_types(cp->action, c, ok, type, rules);
3841 ###### interp exec cases
3842 case Xcond_statement:
3844 struct value v, cnd;
3845 struct type *vtype, *cndtype;
3846 struct casepart *cp;
3847 struct cond_statement *c = cast(cond_statement, e);
3850 interp_exec(c->forpart, NULL);
3853 cnd = interp_exec(c->condpart, &cndtype);
3856 if (!(cndtype == Tnone ||
3857 (cndtype == Tbool && cnd.bool != 0)))
3859 // cnd is Tnone or Tbool, doesn't need to be freed
3861 interp_exec(c->dopart, NULL);
3864 rv = interp_exec(c->thenpart, &rvtype);
3865 if (rvtype != Tnone || !c->dopart)
3867 free_value(rvtype, &rv);
3870 } while (c->dopart);
3872 for (cp = c->casepart; cp; cp = cp->next) {
3873 v = interp_exec(cp->value, &vtype);
3874 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
3875 free_value(vtype, &v);
3876 free_value(cndtype, &cnd);
3877 rv = interp_exec(cp->action, &rvtype);
3880 free_value(vtype, &v);
3882 free_value(cndtype, &cnd);
3884 rv = interp_exec(c->elsepart, &rvtype);
3891 ### Top level structure
3893 All the language elements so far can be used in various places. Now
3894 it is time to clarify what those places are.
3896 At the top level of a file there will be a number of declarations.
3897 Many of the things that can be declared haven't been described yet,
3898 such as functions, procedures, imports, and probably more.
3899 For now there are two sorts of things that can appear at the top
3900 level. They are predefined constants, `struct` types, and the main
3901 program. While the syntax will allow the main program to appear
3902 multiple times, that will trigger an error if it is actually attempted.
3904 The various declarations do not return anything. They store the
3905 various declarations in the parse context.
3907 ###### Parser: grammar
3910 Ocean -> OptNL DeclarationList
3912 ## declare terminals
3919 DeclarationList -> Declaration
3920 | DeclarationList Declaration
3922 Declaration -> ERROR Newlines ${
3924 "error: unhandled parse error", &$1);
3930 ## top level grammar
3932 ### The `const` section
3934 As well as being defined in with the code that uses them, constants
3935 can be declared at the top level. These have full-file scope, so they
3936 are always `InScope`. The value of a top level constant can be given
3937 as an expression, and this is evaluated immediately rather than in the
3938 later interpretation stage. Once we add functions to the language, we
3939 will need rules concern which, if any, can be used to define a top
3942 Constants are defined in a section that starts with the reserved word
3943 `const` and then has a block with a list of assignment statements.
3944 For syntactic consistency, these must use the double-colon syntax to
3945 make it clear that they are constants. Type can also be given: if
3946 not, the type will be determined during analysis, as with other
3949 As the types constants are inserted at the head of a list, printing
3950 them in the same order that they were read is not straight forward.
3951 We take a quadratic approach here and count the number of constants
3952 (variables of depth 0), then count down from there, each time
3953 searching through for the Nth constant for decreasing N.
3955 ###### top level grammar
3959 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
3960 | const { SimpleConstList } Newlines
3961 | const IN OptNL ConstList OUT Newlines
3962 | const SimpleConstList Newlines
3964 ConstList -> ConstList SimpleConstLine
3966 SimpleConstList -> SimpleConstList ; Const
3969 SimpleConstLine -> SimpleConstList Newlines
3970 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
3973 CType -> Type ${ $0 = $<1; }$
3976 Const -> IDENTIFIER :: CType = Expression ${ {
3980 v = var_decl(c, $1.txt);
3982 struct var *var = new_pos(var, $1);
3983 v->where_decl = var;
3988 v = var_ref(c, $1.txt);
3989 tok_err(c, "error: name already declared", &$1);
3990 type_err(c, "info: this is where '%v' was first declared",
3991 v->where_decl, NULL, 0, NULL);
3995 propagate_types($5, c, &ok, $3, 0);
4000 struct value res = interp_exec($5, &v->type);
4001 v->val = val_alloc(v->type, &res);
4005 ###### print const decls
4010 while (target != 0) {
4012 for (v = context.in_scope; v; v=v->in_scope)
4013 if (v->depth == 0) {
4024 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4025 type_print(v->type, stdout);
4027 if (v->type == Tstr)
4029 print_value(v->type, v->val);
4030 if (v->type == Tstr)
4038 ### Finally the whole program.
4040 Somewhat reminiscent of Pascal a (current) Ocean program starts with
4041 the keyword "program" and a list of variable names which are assigned
4042 values from command line arguments. Following this is a `block` which
4043 is the code to execute. Unlike Pascal, constants and other
4044 declarations come *before* the program.
4046 As this is the top level, several things are handled a bit
4048 The whole program is not interpreted by `interp_exec` as that isn't
4049 passed the argument list which the program requires. Similarly type
4050 analysis is a bit more interesting at this level.
4055 ###### top level grammar
4057 DeclareProgram -> Program ${ {
4059 type_err(c, "Program defined a second time",
4068 Program -> program OpenScope Varlist ColonBlock Newlines ${
4071 $0->left = reorder_bilist($<Vl);
4073 var_block_close(c, CloseSequential);
4074 if (c->scope_stack && !c->parse_error) abort();
4077 Varlist -> Varlist ArgDecl ${
4086 ArgDecl -> IDENTIFIER ${ {
4087 struct variable *v = var_decl(c, $1.txt);
4094 ###### print binode cases
4096 do_indent(indent, "program");
4097 for (b2 = cast(binode, b->left); b2; b2 = cast(binode, b2->right)) {
4099 print_exec(b2->left, 0, 0);
4105 print_exec(b->right, indent+1, bracket);
4107 do_indent(indent, "}\n");
4110 ###### propagate binode cases
4111 case Program: abort(); // NOTEST
4113 ###### core functions
4115 static int analyse_prog(struct exec *prog, struct parse_context *c)
4117 struct binode *b = cast(binode, prog);
4124 propagate_types(b->right, c, &ok, Tnone, 0);
4129 for (b = cast(binode, b->left); b; b = cast(binode, b->right)) {
4130 struct var *v = cast(var, b->left);
4131 if (!v->var->type) {
4132 v->var->where_set = b;
4133 v->var->type = Tstr;
4137 b = cast(binode, prog);
4140 propagate_types(b->right, c, &ok, Tnone, 0);
4145 /* Make sure everything is still consistent */
4146 propagate_types(b->right, c, &ok, Tnone, 0);
4150 static void interp_prog(struct exec *prog, char **argv)
4152 struct binode *p = cast(binode, prog);
4159 al = cast(binode, p->left);
4161 struct var *v = cast(var, al->left);
4162 struct value *vl = v->var->val;
4164 if (argv[0] == NULL) {
4165 printf("Not enough args\n");
4168 al = cast(binode, al->right);
4170 free_value(v->var->type, vl);
4172 vl = val_alloc(v->var->type, NULL);
4175 free_value(v->var->type, vl);
4176 vl->str.len = strlen(argv[0]);
4177 vl->str.txt = malloc(vl->str.len);
4178 memcpy(vl->str.txt, argv[0], vl->str.len);
4181 v = interp_exec(p->right, &vtype);
4182 free_value(vtype, &v);
4185 ###### interp binode cases
4186 case Program: abort(); // NOTEST
4188 ## And now to test it out.
4190 Having a language requires having a "hello world" program. I'll
4191 provide a little more than that: a program that prints "Hello world"
4192 finds the GCD of two numbers, prints the first few elements of
4193 Fibonacci, performs a binary search for a number, and a few other
4194 things which will likely grow as the languages grows.
4196 ###### File: oceani.mk
4199 @echo "===== DEMO ====="
4200 ./oceani --section "demo: hello" oceani.mdc 55 33
4206 four ::= 2 + 2 ; five ::= 10/2
4207 const pie ::= "I like Pie";
4208 cake ::= "The cake is"
4217 print "Hello World, what lovely oceans you have!"
4218 print "Are there", five, "?"
4219 print pi, pie, "but", cake
4221 A := $Astr; B := $Bstr
4223 /* When a variable is defined in both branches of an 'if',
4224 * and used afterwards, the variables are merged.
4230 print "Is", A, "bigger than", B,"? ", bigger
4231 /* If a variable is not used after the 'if', no
4232 * merge happens, so types can be different
4235 double:string = "yes"
4236 print A, "is more than twice", B, "?", double
4239 print "double", B, "is", double
4244 if a > 0 and then b > 0:
4250 print "GCD of", A, "and", B,"is", a
4252 print a, "is not positive, cannot calculate GCD"
4254 print b, "is not positive, cannot calculate GCD"
4259 print "Fibonacci:", f1,f2,
4260 then togo = togo - 1
4268 /* Binary search... */
4273 mid := (lo + hi) / 2
4285 print "Yay, I found", target
4287 print "Closest I found was", mid
4292 // "middle square" PRNG. Not particularly good, but one my
4293 // Dad taught me - the first one I ever heard of.
4294 for i:=1; then i = i + 1; while i < size:
4295 n := list[i-1] * list[i-1]
4296 list[i] = (n / 100) % 10 000
4298 print "Before sort:",
4299 for i:=0; then i = i + 1; while i < size:
4303 for i := 1; then i=i+1; while i < size:
4304 for j:=i-1; then j=j-1; while j >= 0:
4305 if list[j] > list[j+1]:
4309 print " After sort:",
4310 for i:=0; then i = i + 1; while i < size:
4314 if 1 == 2 then print "yes"; else print "no"
4318 bob.alive = (bob.name == "Hello")
4319 print "bob", "is" if bob.alive else "isn't", "alive"