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 (*prepare_type)(struct type *type);
433 void (*print)(struct type *type, struct value *val);
434 void (*print_type)(struct type *type, FILE *f);
435 int (*cmp_order)(struct type *t1, struct type *t2,
436 struct value *v1, struct value *v2);
437 int (*cmp_eq)(struct type *t1, struct type *t2,
438 struct value *v1, struct value *v2);
439 void (*dup)(struct type *type, struct value *vold, struct value *vnew);
440 void (*free)(struct type *type, struct value *val);
441 void (*free_type)(struct type *t);
442 long long (*to_int)(struct value *v);
443 double (*to_float)(struct value *v);
444 int (*to_mpq)(mpq_t *q, struct value *v);
453 struct type *typelist;
457 static struct type *find_type(struct parse_context *c, struct text s)
459 struct type *l = c->typelist;
462 text_cmp(l->name, s) != 0)
467 static struct type *add_type(struct parse_context *c, struct text s,
472 n = calloc(1, sizeof(*n));
475 n->next = c->typelist;
480 static void free_type(struct type *t)
482 /* The type is always a reference to something in the
483 * context, so we don't need to free anything.
487 static void free_value(struct type *type, struct value *v)
493 static void type_print(struct type *type, FILE *f)
496 fputs("*unknown*type*", f);
497 else if (type->name.len)
498 fprintf(f, "%.*s", type->name.len, type->name.txt);
499 else if (type->print_type)
500 type->print_type(type, f);
502 fputs("*invalid*type*", f); // NOTEST
505 static void val_init(struct type *type, struct value *val)
507 if (type && type->init)
508 type->init(type, val);
511 static void dup_value(struct type *type,
512 struct value *vold, struct value *vnew)
514 if (type && type->dup)
515 type->dup(type, vold, vnew);
518 static int value_cmp(struct type *tl, struct type *tr,
519 struct value *left, struct value *right)
521 if (tl && tl->cmp_order)
522 return tl->cmp_order(tl, tr, left, right);
523 if (tl && tl->cmp_eq)
524 return tl->cmp_eq(tl, tr, left, right);
528 static void print_value(struct type *type, struct value *v)
530 if (type && type->print)
531 type->print(type, v);
533 printf("*Unknown*"); // NOTEST
536 static struct value *val_alloc(struct type *t, struct value *init)
543 ret = calloc(1, t->size);
545 memcpy(ret, init, t->size);
553 static void free_value(struct type *type, struct value *v);
554 static int type_compat(struct type *require, struct type *have, int rules);
555 static void type_print(struct type *type, FILE *f);
556 static void val_init(struct type *type, struct value *v);
557 static void dup_value(struct type *type,
558 struct value *vold, struct value *vnew);
559 static int value_cmp(struct type *tl, struct type *tr,
560 struct value *left, struct value *right);
561 static void print_value(struct type *type, struct value *v);
563 ###### free context types
565 while (context.typelist) {
566 struct type *t = context.typelist;
568 context.typelist = t->next;
576 Values of the base types can be numbers, which we represent as
577 multi-precision fractions, strings, Booleans and labels. When
578 analysing the program we also need to allow for places where no value
579 is meaningful (type `Tnone`) and where we don't know what type to
580 expect yet (type is `NULL`).
582 Values are never shared, they are always copied when used, and freed
583 when no longer needed.
585 When propagating type information around the program, we need to
586 determine if two types are compatible, where type `NULL` is compatible
587 with anything. There are two special cases with type compatibility,
588 both related to the Conditional Statement which will be described
589 later. In some cases a Boolean can be accepted as well as some other
590 primary type, and in others any type is acceptable except a label (`Vlabel`).
591 A separate function encoding these cases will simplify some code later.
595 int (*compat)(struct type *this, struct type *other);
599 static int type_compat(struct type *require, struct type *have, int rules)
601 if ((rules & Rboolok) && have == Tbool)
603 if ((rules & Rnolabel) && have == Tlabel)
605 if (!require || !have)
609 return require->compat(require, have);
611 return require == have;
616 #include "parse_string.h"
617 #include "parse_number.h"
620 myLDLIBS := libnumber.o libstring.o -lgmp
621 LDLIBS := $(filter-out $(myLDLIBS),$(LDLIBS)) $(myLDLIBS)
623 ###### type union fields
624 enum vtype {Vnone, Vstr, Vnum, Vbool, Vlabel} vtype;
626 ###### value union fields
633 static void _free_value(struct type *type, struct value *v)
637 switch (type->vtype) {
639 case Vstr: free(v->str.txt); break;
640 case Vnum: mpq_clear(v->num); break;
646 ###### value functions
648 static void _val_init(struct type *type, struct value *val)
650 switch(type->vtype) {
651 case Vnone: // NOTEST
654 mpq_init(val->num); break;
656 val->str.txt = malloc(1);
662 case Vlabel: // NOTEST
663 val->label = NULL; // NOTEST
668 static void _dup_value(struct type *type,
669 struct value *vold, struct value *vnew)
671 switch (type->vtype) {
672 case Vnone: // NOTEST
675 vnew->label = vold->label;
678 vnew->bool = vold->bool;
682 mpq_set(vnew->num, vold->num);
685 vnew->str.len = vold->str.len;
686 vnew->str.txt = malloc(vnew->str.len);
687 memcpy(vnew->str.txt, vold->str.txt, vnew->str.len);
692 static int _value_cmp(struct type *tl, struct type *tr,
693 struct value *left, struct value *right)
697 return tl - tr; // NOTEST
699 case Vlabel: cmp = left->label == right->label ? 0 : 1; break;
700 case Vnum: cmp = mpq_cmp(left->num, right->num); break;
701 case Vstr: cmp = text_cmp(left->str, right->str); break;
702 case Vbool: cmp = left->bool - right->bool; break;
703 case Vnone: cmp = 0; // NOTEST
708 static void _print_value(struct type *type, struct value *v)
710 switch (type->vtype) {
711 case Vnone: // NOTEST
712 printf("*no-value*"); break; // NOTEST
713 case Vlabel: // NOTEST
714 printf("*label-%p*", v->label); break; // NOTEST
716 printf("%.*s", v->str.len, v->str.txt); break;
718 printf("%s", v->bool ? "True":"False"); break;
723 mpf_set_q(fl, v->num);
724 gmp_printf("%Fg", fl);
731 static void _free_value(struct type *type, struct value *v);
733 static struct type base_prototype = {
735 .print = _print_value,
736 .cmp_order = _value_cmp,
737 .cmp_eq = _value_cmp,
742 static struct type *Tbool, *Tstr, *Tnum, *Tnone, *Tlabel;
745 static struct type *add_base_type(struct parse_context *c, char *n,
746 enum vtype vt, int size)
748 struct text txt = { n, strlen(n) };
751 t = add_type(c, txt, &base_prototype);
754 t->align = size > sizeof(void*) ? sizeof(void*) : size;
755 if (t->size & (t->align - 1))
756 t->size = (t->size | (t->align - 1)) + 1;
760 ###### context initialization
762 Tbool = add_base_type(&context, "Boolean", Vbool, sizeof(char));
763 Tstr = add_base_type(&context, "string", Vstr, sizeof(struct text));
764 Tnum = add_base_type(&context, "number", Vnum, sizeof(mpq_t));
765 Tnone = add_base_type(&context, "none", Vnone, 0);
766 Tlabel = add_base_type(&context, "label", Vlabel, sizeof(void*));
770 Variables are scoped named values. We store the names in a linked list
771 of "bindings" sorted in lexical order, and use sequential search and
778 struct binding *next; // in lexical order
782 This linked list is stored in the parse context so that "reduce"
783 functions can find or add variables, and so the analysis phase can
784 ensure that every variable gets a type.
788 struct binding *varlist; // In lexical order
792 static struct binding *find_binding(struct parse_context *c, struct text s)
794 struct binding **l = &c->varlist;
799 (cmp = text_cmp((*l)->name, s)) < 0)
803 n = calloc(1, sizeof(*n));
810 Each name can be linked to multiple variables defined in different
811 scopes. Each scope starts where the name is declared and continues
812 until the end of the containing code block. Scopes of a given name
813 cannot nest, so a declaration while a name is in-scope is an error.
815 ###### binding fields
816 struct variable *var;
820 struct variable *previous;
823 struct binding *name;
824 struct exec *where_decl;// where name was declared
825 struct exec *where_set; // where type was set
829 While the naming seems strange, we include local constants in the
830 definition of variables. A name declared `var := value` can
831 subsequently be changed, but a name declared `var ::= value` cannot -
834 ###### variable fields
837 Scopes in parallel branches can be partially merged. More
838 specifically, if a given name is declared in both branches of an
839 if/else then its scope is a candidate for merging. Similarly if
840 every branch of an exhaustive switch (e.g. has an "else" clause)
841 declares a given name, then the scopes from the branches are
842 candidates for merging.
844 Note that names declared inside a loop (which is only parallel to
845 itself) are never visible after the loop. Similarly names defined in
846 scopes which are not parallel, such as those started by `for` and
847 `switch`, are never visible after the scope. Only variables defined in
848 both `then` and `else` (including the implicit then after an `if`, and
849 excluding `then` used with `for`) and in all `case`s and `else` of a
850 `switch` or `while` can be visible beyond the `if`/`switch`/`while`.
852 Labels, which are a bit like variables, follow different rules.
853 Labels are not explicitly declared, but if an undeclared name appears
854 in a context where a label is legal, that effectively declares the
855 name as a label. The declaration remains in force (or in scope) at
856 least to the end of the immediately containing block and conditionally
857 in any larger containing block which does not declare the name in some
858 other way. Importantly, the conditional scope extension happens even
859 if the label is only used in one parallel branch of a conditional --
860 when used in one branch it is treated as having been declared in all
863 Merge candidates are tentatively visible beyond the end of the
864 branching statement which creates them. If the name is used, the
865 merge is affirmed and they become a single variable visible at the
866 outer layer. If not - if it is redeclared first - the merge lapses.
868 To track scopes we have an extra stack, implemented as a linked list,
869 which roughly parallels the parse stack and which is used exclusively
870 for scoping. When a new scope is opened, a new frame is pushed and
871 the child-count of the parent frame is incremented. This child-count
872 is used to distinguish between the first of a set of parallel scopes,
873 in which declared variables must not be in scope, and subsequent
874 branches, whether they may already be conditionally scoped.
876 To push a new frame *before* any code in the frame is parsed, we need a
877 grammar reduction. This is most easily achieved with a grammar
878 element which derives the empty string, and creates the new scope when
879 it is recognised. This can be placed, for example, between a keyword
880 like "if" and the code following it.
884 struct scope *parent;
890 struct scope *scope_stack;
893 static void scope_pop(struct parse_context *c)
895 struct scope *s = c->scope_stack;
897 c->scope_stack = s->parent;
902 static void scope_push(struct parse_context *c)
904 struct scope *s = calloc(1, sizeof(*s));
906 c->scope_stack->child_count += 1;
907 s->parent = c->scope_stack;
915 OpenScope -> ${ scope_push(c); }$
916 ClosePara -> ${ var_block_close(c, CloseParallel); }$
918 Each variable records a scope depth and is in one of four states:
920 - "in scope". This is the case between the declaration of the
921 variable and the end of the containing block, and also between
922 the usage with affirms a merge and the end of that block.
924 The scope depth is not greater than the current parse context scope
925 nest depth. When the block of that depth closes, the state will
926 change. To achieve this, all "in scope" variables are linked
927 together as a stack in nesting order.
929 - "pending". The "in scope" block has closed, but other parallel
930 scopes are still being processed. So far, every parallel block at
931 the same level that has closed has declared the name.
933 The scope depth is the depth of the last parallel block that
934 enclosed the declaration, and that has closed.
936 - "conditionally in scope". The "in scope" block and all parallel
937 scopes have closed, and no further mention of the name has been
938 seen. This state includes a secondary nest depth which records the
939 outermost scope seen since the variable became conditionally in
940 scope. If a use of the name is found, the variable becomes "in
941 scope" and that secondary depth becomes the recorded scope depth.
942 If the name is declared as a new variable, the old variable becomes
943 "out of scope" and the recorded scope depth stays unchanged.
945 - "out of scope". The variable is neither in scope nor conditionally
946 in scope. It is permanently out of scope now and can be removed from
947 the "in scope" stack.
949 ###### variable fields
950 int depth, min_depth;
951 enum { OutScope, PendingScope, CondScope, InScope } scope;
952 struct variable *in_scope;
956 struct variable *in_scope;
958 All variables with the same name are linked together using the
959 'previous' link. Those variable that have been affirmatively merged all
960 have a 'merged' pointer that points to one primary variable - the most
961 recently declared instance. When merging variables, we need to also
962 adjust the 'merged' pointer on any other variables that had previously
963 been merged with the one that will no longer be primary.
965 A variable that is no longer the most recent instance of a name may
966 still have "pending" scope, if it might still be merged with most
967 recent instance. These variables don't really belong in the
968 "in_scope" list, but are not immediately removed when a new instance
969 is found. Instead, they are detected and ignored when considering the
970 list of in_scope names.
972 ###### variable fields
973 struct variable *merged;
977 static void variable_merge(struct variable *primary, struct variable *secondary)
983 primary = primary->merged;
985 for (v = primary->previous; v; v=v->previous)
986 if (v == secondary || v == secondary->merged ||
987 v->merged == secondary ||
988 (v->merged && v->merged == secondary->merged)) {
994 ###### free context vars
996 while (context.varlist) {
997 struct binding *b = context.varlist;
998 struct variable *v = b->var;
999 context.varlist = b->next;
1002 struct variable *t = v;
1005 free_value(t->type, t->val);
1008 // This is a global constant
1009 free_exec(t->where_decl);
1014 #### Manipulating Bindings
1016 When a name is conditionally visible, a new declaration discards the
1017 old binding - the condition lapses. Conversely a usage of the name
1018 affirms the visibility and extends it to the end of the containing
1019 block - i.e. the block that contains both the original declaration and
1020 the latest usage. This is determined from `min_depth`. When a
1021 conditionally visible variable gets affirmed like this, it is also
1022 merged with other conditionally visible variables with the same name.
1024 When we parse a variable declaration we either report an error if the
1025 name is currently bound, or create a new variable at the current nest
1026 depth if the name is unbound or bound to a conditionally scoped or
1027 pending-scope variable. If the previous variable was conditionally
1028 scoped, it and its homonyms becomes out-of-scope.
1030 When we parse a variable reference (including non-declarative assignment
1031 "foo = bar") we report an error if the name is not bound or is bound to
1032 a pending-scope variable; update the scope if the name is bound to a
1033 conditionally scoped variable; or just proceed normally if the named
1034 variable is in scope.
1036 When we exit a scope, any variables bound at this level are either
1037 marked out of scope or pending-scoped, depending on whether the scope
1038 was sequential or parallel. Here a "parallel" scope means the "then"
1039 or "else" part of a conditional, or any "case" or "else" branch of a
1040 switch. Other scopes are "sequential".
1042 When exiting a parallel scope we check if there are any variables that
1043 were previously pending and are still visible. If there are, then
1044 there weren't redeclared in the most recent scope, so they cannot be
1045 merged and must become out-of-scope. If it is not the first of
1046 parallel scopes (based on `child_count`), we check that there was a
1047 previous binding that is still pending-scope. If there isn't, the new
1048 variable must now be out-of-scope.
1050 When exiting a sequential scope that immediately enclosed parallel
1051 scopes, we need to resolve any pending-scope variables. If there was
1052 no `else` clause, and we cannot determine that the `switch` was exhaustive,
1053 we need to mark all pending-scope variable as out-of-scope. Otherwise
1054 all pending-scope variables become conditionally scoped.
1057 enum closetype { CloseSequential, CloseParallel, CloseElse };
1059 ###### ast functions
1061 static struct variable *var_decl(struct parse_context *c, struct text s)
1063 struct binding *b = find_binding(c, s);
1064 struct variable *v = b->var;
1066 switch (v ? v->scope : OutScope) {
1068 /* Caller will report the error */
1072 v && v->scope == CondScope;
1074 v->scope = OutScope;
1078 v = calloc(1, sizeof(*v));
1079 v->previous = b->var;
1082 v->min_depth = v->depth = c->scope_depth;
1084 v->in_scope = c->in_scope;
1090 static struct variable *var_ref(struct parse_context *c, struct text s)
1092 struct binding *b = find_binding(c, s);
1093 struct variable *v = b->var;
1094 struct variable *v2;
1096 switch (v ? v->scope : OutScope) {
1099 /* Caller will report the error */
1102 /* All CondScope variables of this name need to be merged
1103 * and become InScope
1105 v->depth = v->min_depth;
1107 for (v2 = v->previous;
1108 v2 && v2->scope == CondScope;
1110 variable_merge(v, v2);
1118 static void var_block_close(struct parse_context *c, enum closetype ct)
1120 /* Close off all variables that are in_scope */
1121 struct variable *v, **vp, *v2;
1124 for (vp = &c->in_scope;
1125 v = *vp, v && v->depth > c->scope_depth && v->min_depth > c->scope_depth;
1127 if (v->name->var == v) switch (ct) {
1129 case CloseParallel: /* handle PendingScope */
1133 if (c->scope_stack->child_count == 1)
1134 v->scope = PendingScope;
1135 else if (v->previous &&
1136 v->previous->scope == PendingScope)
1137 v->scope = PendingScope;
1138 else if (v->type == Tlabel)
1139 v->scope = PendingScope;
1140 else if (v->name->var == v)
1141 v->scope = OutScope;
1142 if (ct == CloseElse) {
1143 /* All Pending variables with this name
1144 * are now Conditional */
1146 v2 && v2->scope == PendingScope;
1148 v2->scope = CondScope;
1153 v2 && v2->scope == PendingScope;
1155 if (v2->type != Tlabel)
1156 v2->scope = OutScope;
1158 case OutScope: break;
1161 case CloseSequential:
1162 if (v->type == Tlabel)
1163 v->scope = PendingScope;
1166 v->scope = OutScope;
1169 /* There was no 'else', so we can only become
1170 * conditional if we know the cases were exhaustive,
1171 * and that doesn't mean anything yet.
1172 * So only labels become conditional..
1175 v2 && v2->scope == PendingScope;
1177 if (v2->type == Tlabel) {
1178 v2->scope = CondScope;
1179 v2->min_depth = c->scope_depth;
1181 v2->scope = OutScope;
1184 case OutScope: break;
1188 if (v->scope == OutScope || v->name->var != v)
1197 Executables can be lots of different things. In many cases an
1198 executable is just an operation combined with one or two other
1199 executables. This allows for expressions and lists etc. Other times an
1200 executable is something quite specific like a constant or variable name.
1201 So we define a `struct exec` to be a general executable with a type, and
1202 a `struct binode` which is a subclass of `exec`, forms a node in a
1203 binary tree, and holds an operation. There will be other subclasses,
1204 and to access these we need to be able to `cast` the `exec` into the
1205 various other types. The first field in any `struct exec` is the type
1206 from the `exec_types` enum.
1209 #define cast(structname, pointer) ({ \
1210 const typeof( ((struct structname *)0)->type) *__mptr = &(pointer)->type; \
1211 if (__mptr && *__mptr != X##structname) abort(); \
1212 (struct structname *)( (char *)__mptr);})
1214 #define new(structname) ({ \
1215 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1216 __ptr->type = X##structname; \
1217 __ptr->line = -1; __ptr->column = -1; \
1220 #define new_pos(structname, token) ({ \
1221 struct structname *__ptr = ((struct structname *)calloc(1,sizeof(struct structname))); \
1222 __ptr->type = X##structname; \
1223 __ptr->line = token.line; __ptr->column = token.col; \
1232 enum exec_types type;
1240 struct exec *left, *right;
1243 ###### ast functions
1245 static int __fput_loc(struct exec *loc, FILE *f)
1249 if (loc->line >= 0) {
1250 fprintf(f, "%d:%d: ", loc->line, loc->column);
1253 if (loc->type == Xbinode)
1254 return __fput_loc(cast(binode,loc)->left, f) ||
1255 __fput_loc(cast(binode,loc)->right, f);
1258 static void fput_loc(struct exec *loc, FILE *f)
1260 if (!__fput_loc(loc, f))
1261 fprintf(f, "??:??: "); // NOTEST
1264 Each different type of `exec` node needs a number of functions defined,
1265 a bit like methods. We must be able to free it, print it, analyse it
1266 and execute it. Once we have specific `exec` types we will need to
1267 parse them too. Let's take this a bit more slowly.
1271 The parser generator requires a `free_foo` function for each struct
1272 that stores attributes and they will often be `exec`s and subtypes
1273 there-of. So we need `free_exec` which can handle all the subtypes,
1274 and we need `free_binode`.
1276 ###### ast functions
1278 static void free_binode(struct binode *b)
1283 free_exec(b->right);
1287 ###### core functions
1288 static void free_exec(struct exec *e)
1297 ###### forward decls
1299 static void free_exec(struct exec *e);
1301 ###### free exec cases
1302 case Xbinode: free_binode(cast(binode, e)); break;
1306 Printing an `exec` requires that we know the current indent level for
1307 printing line-oriented components. As will become clear later, we
1308 also want to know what sort of bracketing to use.
1310 ###### ast functions
1312 static void do_indent(int i, char *str)
1319 ###### core functions
1320 static void print_binode(struct binode *b, int indent, int bracket)
1324 ## print binode cases
1328 static void print_exec(struct exec *e, int indent, int bracket)
1334 print_binode(cast(binode, e), indent, bracket); break;
1339 ###### forward decls
1341 static void print_exec(struct exec *e, int indent, int bracket);
1345 As discussed, analysis involves propagating type requirements around the
1346 program and looking for errors.
1348 So `propagate_types` is passed an expected type (being a `struct type`
1349 pointer together with some `val_rules` flags) that the `exec` is
1350 expected to return, and returns the type that it does return, either
1351 of which can be `NULL` signifying "unknown". An `ok` flag is passed
1352 by reference. It is set to `0` when an error is found, and `2` when
1353 any change is made. If it remains unchanged at `1`, then no more
1354 propagation is needed.
1358 enum val_rules {Rnolabel = 1<<0, Rboolok = 1<<1, Rnoconstant = 2<<1};
1362 if (rules & Rnolabel)
1363 fputs(" (labels not permitted)", stderr);
1366 ###### core functions
1368 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1369 struct type *type, int rules);
1370 static struct type *__propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1371 struct type *type, int rules)
1378 switch (prog->type) {
1381 struct binode *b = cast(binode, prog);
1383 ## propagate binode cases
1387 ## propagate exec cases
1392 static struct type *propagate_types(struct exec *prog, struct parse_context *c, int *ok,
1393 struct type *type, int rules)
1395 struct type *ret = __propagate_types(prog, c, ok, type, rules);
1404 Interpreting an `exec` doesn't require anything but the `exec`. State
1405 is stored in variables and each variable will be directly linked from
1406 within the `exec` tree. The exception to this is the whole `program`
1407 which needs to look at command line arguments. The `program` will be
1408 interpreted separately.
1410 Each `exec` can return a value combined with a type in `struct lrval`.
1411 The type may be `Tnone` but must be non-NULL. Some `exec`s will return
1412 the location of a value, which can be updated, in `lval`. Others will
1413 set `lval` to NULL indicating that there is a value of appropriate type
1417 ###### core functions
1421 struct value rval, *lval;
1424 static struct lrval _interp_exec(struct exec *e);
1426 static struct value interp_exec(struct exec *e, struct type **typeret)
1428 struct lrval ret = _interp_exec(e);
1430 if (!ret.type) abort();
1432 *typeret = ret.type;
1434 dup_value(ret.type, ret.lval, &ret.rval);
1438 static struct value *linterp_exec(struct exec *e, struct type **typeret)
1440 struct lrval ret = _interp_exec(e);
1443 *typeret = ret.type;
1445 free_value(ret.type, &ret.rval);
1449 static struct lrval _interp_exec(struct exec *e)
1452 struct value rv = {}, *lrv = NULL;
1453 struct type *rvtype;
1455 rvtype = ret.type = Tnone;
1465 struct binode *b = cast(binode, e);
1466 struct value left, right, *lleft;
1467 struct type *ltype, *rtype;
1468 ltype = rtype = Tnone;
1470 ## interp binode cases
1472 free_value(ltype, &left);
1473 free_value(rtype, &right);
1476 ## interp exec cases
1486 Now that we have the shape of the interpreter in place we can add some
1487 complex types and connected them in to the data structures and the
1488 different phases of parse, analyse, print, interpret.
1490 Thus far we have arrays and structs.
1494 Arrays can be declared by giving a size and a type, as `[size]type' so
1495 `freq:[26]number` declares `freq` to be an array of 26 numbers. The
1496 size can be either a literal number, or a named constant. Some day an
1497 arbitrary expression will be supported.
1499 Arrays cannot be assigned. When pointers are introduced we will also
1500 introduce array slices which can refer to part or all of an array -
1501 the assignment syntax will create a slice. For now, an array can only
1502 ever be referenced by the name it is declared with. It is likely that
1503 a "`copy`" primitive will eventually be define which can be used to
1504 make a copy of an array with controllable recursive depth.
1506 ###### type union fields
1510 struct variable *vsize;
1511 struct type *member;
1514 ###### value functions
1516 static void array_prepare_type(struct type *type)
1519 if (!type->array.vsize)
1523 mpz_tdiv_q(q, mpq_numref(type->array.vsize->val->num),
1524 mpq_denref(type->array.vsize->val->num));
1525 type->array.size = mpz_get_si(q);
1528 type->size = type->array.size * type->array.member->size;
1529 type->align = type->array.member->align;
1532 static void array_init(struct type *type, struct value *val)
1538 for (i = 0; i < type->array.size; i++) {
1540 v = (void*)val->ptr + i * type->array.member->size;
1541 val_init(type->array.member, v);
1545 static void array_free(struct type *type, struct value *val)
1549 for (i = 0; i < type->array.size; i++) {
1551 v = (void*)val->ptr + i * type->array.member->size;
1552 free_value(type->array.member, v);
1556 static int array_compat(struct type *require, struct type *have)
1558 if (have->compat != require->compat)
1560 /* Both are arrays, so we can look at details */
1561 if (!type_compat(require->array.member, have->array.member, 0))
1563 if (require->array.vsize == NULL && have->array.vsize == NULL)
1564 return require->array.size == have->array.size;
1566 return require->array.vsize == have->array.vsize;
1569 static void array_print_type(struct type *type, FILE *f)
1572 if (type->array.vsize) {
1573 struct binding *b = type->array.vsize->name;
1574 fprintf(f, "%.*s]", b->name.len, b->name.txt);
1576 fprintf(f, "%d]", type->array.size);
1577 type_print(type->array.member, f);
1580 static struct type array_prototype = {
1582 .prepare_type = array_prepare_type,
1583 .print_type = array_print_type,
1584 .compat = array_compat,
1588 ###### declare terminals
1593 | [ NUMBER ] Type ${ {
1596 struct text noname = { "", 0 };
1599 $0 = t = add_type(c, noname, &array_prototype);
1600 t->array.member = $<4;
1601 t->array.vsize = NULL;
1602 if (number_parse(num, tail, $2.txt) == 0)
1603 tok_err(c, "error: unrecognised number", &$2);
1605 tok_err(c, "error: unsupported number suffix", &$2);
1607 t->array.size = mpz_get_ui(mpq_numref(num));
1608 if (mpz_cmp_ui(mpq_denref(num), 1) != 0) {
1609 tok_err(c, "error: array size must be an integer",
1611 } else if (mpz_cmp_ui(mpq_numref(num), 1UL << 30) >= 0)
1612 tok_err(c, "error: array size is too large",
1616 t->size = t->array.size * t->array.member->size;
1617 t->align = t->array.member->align;
1620 | [ IDENTIFIER ] Type ${ {
1621 struct variable *v = var_ref(c, $2.txt);
1622 struct text noname = { "", 0 };
1625 tok_err(c, "error: name undeclared", &$2);
1626 else if (!v->constant)
1627 tok_err(c, "error: array size must be a constant", &$2);
1629 $0 = add_type(c, noname, &array_prototype);
1630 $0->array.member = $<4;
1632 $0->array.vsize = v;
1638 ###### variable grammar
1640 | Variable [ Expression ] ${ {
1641 struct binode *b = new(binode);
1648 ###### print binode cases
1650 print_exec(b->left, -1, bracket);
1652 print_exec(b->right, -1, bracket);
1656 ###### propagate binode cases
1658 /* left must be an array, right must be a number,
1659 * result is the member type of the array
1661 propagate_types(b->right, c, ok, Tnum, 0);
1662 t = propagate_types(b->left, c, ok, NULL, rules & Rnoconstant);
1663 if (!t || t->compat != array_compat) {
1664 type_err(c, "error: %1 cannot be indexed", prog, t, 0, NULL);
1667 if (!type_compat(type, t->array.member, rules)) {
1668 type_err(c, "error: have %1 but need %2", prog,
1669 t->array.member, rules, type);
1671 return t->array.member;
1675 ###### interp binode cases
1680 lleft = linterp_exec(b->left, <ype);
1681 right = interp_exec(b->right, &rtype);
1683 mpz_tdiv_q(q, mpq_numref(right.num), mpq_denref(right.num));
1687 rvtype = ltype->array.member;
1688 if (i >= 0 && i < ltype->array.size)
1689 lrv = (void*)lleft + i * rvtype->size;
1691 val_init(ltype->array.member, &rv);
1698 A `struct` is a data-type that contains one or more other data-types.
1699 It differs from an array in that each member can be of a different
1700 type, and they are accessed by name rather than by number. Thus you
1701 cannot choose an element by calculation, you need to know what you
1704 The language makes no promises about how a given structure will be
1705 stored in memory - it is free to rearrange fields to suit whatever
1706 criteria seems important.
1708 Structs are declared separately from program code - they cannot be
1709 declared in-line in a variable declaration like arrays can. A struct
1710 is given a name and this name is used to identify the type - the name
1711 is not prefixed by the word `struct` as it would be in C.
1713 Structs are only treated as the same if they have the same name.
1714 Simply having the same fields in the same order is not enough. This
1715 might change once we can create structure initializers from a list of
1718 Each component datum is identified much like a variable is declared,
1719 with a name, one or two colons, and a type. The type cannot be omitted
1720 as there is no opportunity to deduce the type from usage. An initial
1721 value can be given following an equals sign, so
1723 ##### Example: a struct type
1729 would declare a type called "complex" which has two number fields,
1730 each initialised to zero.
1732 Struct will need to be declared separately from the code that uses
1733 them, so we will need to be able to print out the declaration of a
1734 struct when reprinting the whole program. So a `print_type_decl` type
1735 function will be needed.
1737 ###### type union fields
1749 ###### type functions
1750 void (*print_type_decl)(struct type *type, FILE *f);
1752 ###### value functions
1754 static void structure_init(struct type *type, struct value *val)
1758 for (i = 0; i < type->structure.nfields; i++) {
1760 v = (void*) val->ptr + type->structure.fields[i].offset;
1761 if (type->structure.fields[i].init)
1762 dup_value(type->structure.fields[i].type,
1763 type->structure.fields[i].init,
1766 val_init(type->structure.fields[i].type, v);
1770 static void structure_free(struct type *type, struct value *val)
1774 for (i = 0; i < type->structure.nfields; i++) {
1776 v = (void*)val->ptr + type->structure.fields[i].offset;
1777 free_value(type->structure.fields[i].type, v);
1781 static void structure_free_type(struct type *t)
1784 for (i = 0; i < t->structure.nfields; i++)
1785 if (t->structure.fields[i].init) {
1786 free_value(t->structure.fields[i].type,
1787 t->structure.fields[i].init);
1788 free(t->structure.fields[i].init);
1790 free(t->structure.fields);
1793 static struct type structure_prototype = {
1794 .init = structure_init,
1795 .free = structure_free,
1796 .free_type = structure_free_type,
1797 .print_type_decl = structure_print_type,
1811 ###### free exec cases
1813 free_exec(cast(fieldref, e)->left);
1817 ###### declare terminals
1820 ###### variable grammar
1822 | Variable . IDENTIFIER ${ {
1823 struct fieldref *fr = new_pos(fieldref, $2);
1830 ###### print exec cases
1834 struct fieldref *f = cast(fieldref, e);
1835 print_exec(f->left, -1, bracket);
1836 printf(".%.*s", f->name.len, f->name.txt);
1840 ###### ast functions
1841 static int find_struct_index(struct type *type, struct text field)
1844 for (i = 0; i < type->structure.nfields; i++)
1845 if (text_cmp(type->structure.fields[i].name, field) == 0)
1850 ###### propagate exec cases
1854 struct fieldref *f = cast(fieldref, prog);
1855 struct type *st = propagate_types(f->left, c, ok, NULL, 0);
1858 type_err(c, "error: unknown type for field access", f->left,
1860 else if (st->init != structure_init)
1861 type_err(c, "error: field reference attempted on %1, not a struct",
1862 f->left, st, 0, NULL);
1863 else if (f->index == -2) {
1864 f->index = find_struct_index(st, f->name);
1866 type_err(c, "error: cannot find requested field in %1",
1867 f->left, st, 0, NULL);
1869 if (f->index >= 0) {
1870 struct type *ft = st->structure.fields[f->index].type;
1871 if (!type_compat(type, ft, rules))
1872 type_err(c, "error: have %1 but need %2", prog,
1879 ###### interp exec cases
1882 struct fieldref *f = cast(fieldref, e);
1884 struct value *lleft = linterp_exec(f->left, <ype);
1885 lrv = (void*)lleft->ptr + ltype->structure.fields[f->index].offset;
1886 rvtype = ltype->structure.fields[f->index].type;
1892 struct fieldlist *prev;
1896 ###### ast functions
1897 static void free_fieldlist(struct fieldlist *f)
1901 free_fieldlist(f->prev);
1903 free_value(f->f.type, f->f.init);
1909 ###### top level grammar
1910 DeclareStruct -> struct IDENTIFIER FieldBlock Newlines ${ {
1912 add_type(c, $2.txt, &structure_prototype);
1914 struct fieldlist *f;
1916 for (f = $3; f; f=f->prev)
1919 t->structure.nfields = cnt;
1920 t->structure.fields = calloc(cnt, sizeof(struct field));
1923 int a = f->f.type->align;
1925 t->structure.fields[cnt] = f->f;
1926 if (t->size & (a-1))
1927 t->size = (t->size | (a-1)) + 1;
1928 t->structure.fields[cnt].offset = t->size;
1929 t->size += ((f->f.type->size - 1) | (a-1)) + 1;
1938 FieldBlock -> { IN OptNL FieldLines OUT OptNL } ${ $0 = $<FL; }$
1939 | { SimpleFieldList } ${ $0 = $<SFL; }$
1940 | IN OptNL FieldLines OUT ${ $0 = $<FL; }$
1941 | SimpleFieldList EOL ${ $0 = $<SFL; }$
1943 FieldLines -> SimpleFieldList Newlines ${ $0 = $<SFL; }$
1944 | FieldLines SimpleFieldList Newlines ${
1949 SimpleFieldList -> Field ${ $0 = $<F; }$
1950 | SimpleFieldList ; Field ${
1954 | SimpleFieldList ; ${
1957 | ERROR ${ tok_err(c, "Syntax error in struct field", &$1); }$
1959 Field -> IDENTIFIER : Type = Expression ${ {
1962 $0 = calloc(1, sizeof(struct fieldlist));
1963 $0->f.name = $1.txt;
1968 propagate_types($<5, c, &ok, $3, 0);
1973 struct value vl = interp_exec($5, NULL);
1974 $0->f.init = val_alloc($0->f.type, &vl);
1977 | IDENTIFIER : Type ${
1978 $0 = calloc(1, sizeof(struct fieldlist));
1979 $0->f.name = $1.txt;
1981 $0->f.init = val_alloc($0->f.type, NULL);
1984 ###### forward decls
1985 static void structure_print_type(struct type *t, FILE *f);
1987 ###### value functions
1988 static void structure_print_type(struct type *t, FILE *f)
1992 fprintf(f, "struct %.*s\n", t->name.len, t->name.txt);
1994 for (i = 0; i < t->structure.nfields; i++) {
1995 struct field *fl = t->structure.fields + i;
1996 fprintf(f, " %.*s : ", fl->name.len, fl->name.txt);
1997 type_print(fl->type, f);
1998 if (fl->type->print && fl->init) {
2000 if (fl->type == Tstr)
2002 print_value(fl->type, fl->init);
2003 if (fl->type == Tstr)
2010 ###### print type decls
2015 while (target != 0) {
2017 for (t = context.typelist; t ; t=t->next)
2018 if (t->print_type_decl) {
2027 t->print_type_decl(t, stdout);
2033 ## Executables: the elements of code
2035 Each code element needs to be parsed, printed, analysed,
2036 interpreted, and freed. There are several, so let's just start with
2037 the easy ones and work our way up.
2041 We have already met values as separate objects. When manifest
2042 constants appear in the program text, that must result in an executable
2043 which has a constant value. So the `val` structure embeds a value in
2056 ###### ast functions
2057 struct val *new_val(struct type *T, struct token tk)
2059 struct val *v = new_pos(val, tk);
2070 $0 = new_val(Tbool, $1);
2074 $0 = new_val(Tbool, $1);
2078 $0 = new_val(Tnum, $1);
2081 if (number_parse($0->val.num, tail, $1.txt) == 0)
2082 mpq_init($0->val.num);
2084 tok_err(c, "error: unsupported number suffix",
2089 $0 = new_val(Tstr, $1);
2092 string_parse(&$1, '\\', &$0->val.str, tail);
2094 tok_err(c, "error: unsupported string suffix",
2099 $0 = new_val(Tstr, $1);
2102 string_parse(&$1, '\\', &$0->val.str, tail);
2104 tok_err(c, "error: unsupported string suffix",
2109 ###### print exec cases
2112 struct val *v = cast(val, e);
2113 if (v->vtype == Tstr)
2115 print_value(v->vtype, &v->val);
2116 if (v->vtype == Tstr)
2121 ###### propagate exec cases
2124 struct val *val = cast(val, prog);
2125 if (!type_compat(type, val->vtype, rules))
2126 type_err(c, "error: expected %1%r found %2",
2127 prog, type, rules, val->vtype);
2131 ###### interp exec cases
2133 rvtype = cast(val, e)->vtype;
2134 dup_value(rvtype, &cast(val, e)->val, &rv);
2137 ###### ast functions
2138 static void free_val(struct val *v)
2141 free_value(v->vtype, &v->val);
2145 ###### free exec cases
2146 case Xval: free_val(cast(val, e)); break;
2148 ###### ast functions
2149 // Move all nodes from 'b' to 'rv', reversing their order.
2150 // In 'b' 'left' is a list, and 'right' is the last node.
2151 // In 'rv', left' is the first node and 'right' is a list.
2152 static struct binode *reorder_bilist(struct binode *b)
2154 struct binode *rv = NULL;
2157 struct exec *t = b->right;
2161 b = cast(binode, b->left);
2171 Just as we used a `val` to wrap a value into an `exec`, we similarly
2172 need a `var` to wrap a `variable` into an exec. While each `val`
2173 contained a copy of the value, each `var` holds a link to the variable
2174 because it really is the same variable no matter where it appears.
2175 When a variable is used, we need to remember to follow the `->merged`
2176 link to find the primary instance.
2184 struct variable *var;
2192 VariableDecl -> IDENTIFIER : ${ {
2193 struct variable *v = var_decl(c, $1.txt);
2194 $0 = new_pos(var, $1);
2199 v = var_ref(c, $1.txt);
2201 type_err(c, "error: variable '%v' redeclared",
2203 type_err(c, "info: this is where '%v' was first declared",
2204 v->where_decl, NULL, 0, NULL);
2207 | IDENTIFIER :: ${ {
2208 struct variable *v = var_decl(c, $1.txt);
2209 $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);
2233 v = var_ref(c, $1.txt);
2235 type_err(c, "error: variable '%v' redeclared",
2237 type_err(c, "info: this is where '%v' was first declared",
2238 v->where_decl, NULL, 0, NULL);
2241 | IDENTIFIER :: Type ${ {
2242 struct variable *v = var_decl(c, $1.txt);
2243 $0 = new_pos(var, $1);
2252 v = var_ref(c, $1.txt);
2254 type_err(c, "error: variable '%v' redeclared",
2256 type_err(c, "info: this is where '%v' was first declared",
2257 v->where_decl, NULL, 0, NULL);
2262 Variable -> IDENTIFIER ${ {
2263 struct variable *v = var_ref(c, $1.txt);
2264 $0 = new_pos(var, $1);
2266 /* This might be a label - allocate a var just in case */
2267 v = var_decl(c, $1.txt);
2275 cast(var, $0)->var = v;
2280 Type -> IDENTIFIER ${
2281 $0 = find_type(c, $1.txt);
2284 "error: undefined type", &$1);
2291 ###### print exec cases
2294 struct var *v = cast(var, e);
2296 struct binding *b = v->var->name;
2297 printf("%.*s", b->name.len, b->name.txt);
2304 if (loc->type == Xvar) {
2305 struct var *v = cast(var, loc);
2307 struct binding *b = v->var->name;
2308 fprintf(stderr, "%.*s", b->name.len, b->name.txt);
2310 fputs("???", stderr); // NOTEST
2312 fputs("NOTVAR", stderr); // NOTEST
2315 ###### propagate exec cases
2319 struct var *var = cast(var, prog);
2320 struct variable *v = var->var;
2322 type_err(c, "%d:BUG: no variable!!", prog, NULL, 0, NULL); // NOTEST
2323 return Tnone; // NOTEST
2327 if (v->constant && (rules & Rnoconstant)) {
2328 type_err(c, "error: Cannot assign to a constant: %v",
2329 prog, NULL, 0, NULL);
2330 type_err(c, "info: name was defined as a constant here",
2331 v->where_decl, NULL, 0, NULL);
2334 if (v->type == Tnone && v->where_decl == prog)
2335 type_err(c, "error: variable used but not declared: %v",
2336 prog, NULL, 0, NULL);
2337 if (v->type == NULL) {
2338 if (type && *ok != 0) {
2341 v->where_set = prog;
2346 if (!type_compat(type, v->type, rules)) {
2347 type_err(c, "error: expected %1%r but variable '%v' is %2", prog,
2348 type, rules, v->type);
2349 type_err(c, "info: this is where '%v' was set to %1", v->where_set,
2350 v->type, rules, NULL);
2357 ###### interp exec cases
2360 struct var *var = cast(var, e);
2361 struct variable *v = var->var;
2370 ###### ast functions
2372 static void free_var(struct var *v)
2377 ###### free exec cases
2378 case Xvar: free_var(cast(var, e)); break;
2380 ### Expressions: Conditional
2382 Our first user of the `binode` will be conditional expressions, which
2383 is a bit odd as they actually have three components. That will be
2384 handled by having 2 binodes for each expression. The conditional
2385 expression is the lowest precedence operator which is why we define it
2386 first - to start the precedence list.
2388 Conditional expressions are of the form "value `if` condition `else`
2389 other_value". They associate to the right, so everything to the right
2390 of `else` is part of an else value, while only a higher-precedence to
2391 the left of `if` is the if values. Between `if` and `else` there is no
2392 room for ambiguity, so a full conditional expression is allowed in
2404 Expression -> Expression if Expression else Expression $$ifelse ${ {
2405 struct binode *b1 = new(binode);
2406 struct binode *b2 = new(binode);
2415 ## expression grammar
2417 ###### print binode cases
2420 b2 = cast(binode, b->right);
2421 if (bracket) printf("(");
2422 print_exec(b2->left, -1, bracket);
2424 print_exec(b->left, -1, bracket);
2426 print_exec(b2->right, -1, bracket);
2427 if (bracket) printf(")");
2430 ###### propagate binode cases
2433 /* cond must be Tbool, others must match */
2434 struct binode *b2 = cast(binode, b->right);
2437 propagate_types(b->left, c, ok, Tbool, 0);
2438 t = propagate_types(b2->left, c, ok, type, Rnolabel);
2439 t2 = propagate_types(b2->right, c, ok, type ?: t, Rnolabel);
2443 ###### interp binode cases
2446 struct binode *b2 = cast(binode, b->right);
2447 left = interp_exec(b->left, <ype);
2449 rv = interp_exec(b2->left, &rvtype);
2451 rv = interp_exec(b2->right, &rvtype);
2455 ### Expressions: Boolean
2457 The next class of expressions to use the `binode` will be Boolean
2458 expressions. "`and then`" and "`or else`" are similar to `and` and `or`
2459 have same corresponding precendence. The difference is that they don't
2460 evaluate the second expression if not necessary.
2469 ###### expr precedence
2474 ###### expression grammar
2475 | Expression or Expression ${ {
2476 struct binode *b = new(binode);
2482 | Expression or else Expression ${ {
2483 struct binode *b = new(binode);
2490 | Expression and Expression ${ {
2491 struct binode *b = new(binode);
2497 | Expression and then Expression ${ {
2498 struct binode *b = new(binode);
2505 | not Expression ${ {
2506 struct binode *b = new(binode);
2512 ###### print binode cases
2514 if (bracket) printf("(");
2515 print_exec(b->left, -1, bracket);
2517 print_exec(b->right, -1, bracket);
2518 if (bracket) printf(")");
2521 if (bracket) printf("(");
2522 print_exec(b->left, -1, bracket);
2523 printf(" and then ");
2524 print_exec(b->right, -1, bracket);
2525 if (bracket) printf(")");
2528 if (bracket) printf("(");
2529 print_exec(b->left, -1, bracket);
2531 print_exec(b->right, -1, bracket);
2532 if (bracket) printf(")");
2535 if (bracket) printf("(");
2536 print_exec(b->left, -1, bracket);
2537 printf(" or else ");
2538 print_exec(b->right, -1, bracket);
2539 if (bracket) printf(")");
2542 if (bracket) printf("(");
2544 print_exec(b->right, -1, bracket);
2545 if (bracket) printf(")");
2548 ###### propagate binode cases
2554 /* both must be Tbool, result is Tbool */
2555 propagate_types(b->left, c, ok, Tbool, 0);
2556 propagate_types(b->right, c, ok, Tbool, 0);
2557 if (type && type != Tbool)
2558 type_err(c, "error: %1 operation found where %2 expected", prog,
2562 ###### interp binode cases
2564 rv = interp_exec(b->left, &rvtype);
2565 right = interp_exec(b->right, &rtype);
2566 rv.bool = rv.bool && right.bool;
2569 rv = interp_exec(b->left, &rvtype);
2571 rv = interp_exec(b->right, NULL);
2574 rv = interp_exec(b->left, &rvtype);
2575 right = interp_exec(b->right, &rtype);
2576 rv.bool = rv.bool || right.bool;
2579 rv = interp_exec(b->left, &rvtype);
2581 rv = interp_exec(b->right, NULL);
2584 rv = interp_exec(b->right, &rvtype);
2588 ### Expressions: Comparison
2590 Of slightly higher precedence that Boolean expressions are Comparisons.
2591 A comparison takes arguments of any comparable type, but the two types
2594 To simplify the parsing we introduce an `eop` which can record an
2595 expression operator, and the `CMPop` non-terminal will match one of them.
2602 ###### ast functions
2603 static void free_eop(struct eop *e)
2617 ###### expr precedence
2618 $LEFT < > <= >= == != CMPop
2620 ###### expression grammar
2621 | Expression CMPop Expression ${ {
2622 struct binode *b = new(binode);
2632 CMPop -> < ${ $0.op = Less; }$
2633 | > ${ $0.op = Gtr; }$
2634 | <= ${ $0.op = LessEq; }$
2635 | >= ${ $0.op = GtrEq; }$
2636 | == ${ $0.op = Eql; }$
2637 | != ${ $0.op = NEql; }$
2639 ###### print binode cases
2647 if (bracket) printf("(");
2648 print_exec(b->left, -1, bracket);
2650 case Less: printf(" < "); break;
2651 case LessEq: printf(" <= "); break;
2652 case Gtr: printf(" > "); break;
2653 case GtrEq: printf(" >= "); break;
2654 case Eql: printf(" == "); break;
2655 case NEql: printf(" != "); break;
2656 default: abort(); // NOTEST
2658 print_exec(b->right, -1, bracket);
2659 if (bracket) printf(")");
2662 ###### propagate binode cases
2669 /* Both must match but not be labels, result is Tbool */
2670 t = propagate_types(b->left, c, ok, NULL, Rnolabel);
2672 propagate_types(b->right, c, ok, t, 0);
2674 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
2676 t = propagate_types(b->left, c, ok, t, 0);
2678 if (!type_compat(type, Tbool, 0))
2679 type_err(c, "error: Comparison returns %1 but %2 expected", prog,
2680 Tbool, rules, type);
2683 ###### interp binode cases
2692 left = interp_exec(b->left, <ype);
2693 right = interp_exec(b->right, &rtype);
2694 cmp = value_cmp(ltype, rtype, &left, &right);
2697 case Less: rv.bool = cmp < 0; break;
2698 case LessEq: rv.bool = cmp <= 0; break;
2699 case Gtr: rv.bool = cmp > 0; break;
2700 case GtrEq: rv.bool = cmp >= 0; break;
2701 case Eql: rv.bool = cmp == 0; break;
2702 case NEql: rv.bool = cmp != 0; break;
2703 default: rv.bool = 0; break; // NOTEST
2708 ### Expressions: The rest
2710 The remaining expressions with the highest precedence are arithmetic,
2711 string concatenation, and string conversion. String concatenation
2712 (`++`) has the same precedence as multiplication and division, but lower
2715 String conversion is a temporary feature until I get a better type
2716 system. `$` is a prefix operator which expects a string and returns
2719 `+` and `-` are both infix and prefix operations (where they are
2720 absolute value and negation). These have different operator names.
2722 We also have a 'Bracket' operator which records where parentheses were
2723 found. This makes it easy to reproduce these when printing. Possibly I
2724 should only insert brackets were needed for precedence.
2734 ###### expr precedence
2740 ###### expression grammar
2741 | Expression Eop Expression ${ {
2742 struct binode *b = new(binode);
2749 | Expression Top Expression ${ {
2750 struct binode *b = new(binode);
2757 | ( Expression ) ${ {
2758 struct binode *b = new_pos(binode, $1);
2763 | Uop Expression ${ {
2764 struct binode *b = new(binode);
2769 | Value ${ $0 = $<1; }$
2770 | Variable ${ $0 = $<1; }$
2773 Eop -> + ${ $0.op = Plus; }$
2774 | - ${ $0.op = Minus; }$
2776 Uop -> + ${ $0.op = Absolute; }$
2777 | - ${ $0.op = Negate; }$
2778 | $ ${ $0.op = StringConv; }$
2780 Top -> * ${ $0.op = Times; }$
2781 | / ${ $0.op = Divide; }$
2782 | % ${ $0.op = Rem; }$
2783 | ++ ${ $0.op = Concat; }$
2785 ###### print binode cases
2792 if (bracket) printf("(");
2793 print_exec(b->left, indent, bracket);
2795 case Plus: fputs(" + ", stdout); break;
2796 case Minus: fputs(" - ", stdout); break;
2797 case Times: fputs(" * ", stdout); break;
2798 case Divide: fputs(" / ", stdout); break;
2799 case Rem: fputs(" % ", stdout); break;
2800 case Concat: fputs(" ++ ", stdout); break;
2801 default: abort(); // NOTEST
2803 print_exec(b->right, indent, bracket);
2804 if (bracket) printf(")");
2809 if (bracket) printf("(");
2811 case Absolute: fputs("+", stdout); break;
2812 case Negate: fputs("-", stdout); break;
2813 case StringConv: fputs("$", stdout); break;
2814 default: abort(); // NOTEST
2816 print_exec(b->right, indent, bracket);
2817 if (bracket) printf(")");
2821 print_exec(b->right, indent, bracket);
2825 ###### propagate binode cases
2831 /* both must be numbers, result is Tnum */
2834 /* as propagate_types ignores a NULL,
2835 * unary ops fit here too */
2836 propagate_types(b->left, c, ok, Tnum, 0);
2837 propagate_types(b->right, c, ok, Tnum, 0);
2838 if (!type_compat(type, Tnum, 0))
2839 type_err(c, "error: Arithmetic returns %1 but %2 expected", prog,
2844 /* both must be Tstr, result is Tstr */
2845 propagate_types(b->left, c, ok, Tstr, 0);
2846 propagate_types(b->right, c, ok, Tstr, 0);
2847 if (!type_compat(type, Tstr, 0))
2848 type_err(c, "error: Concat returns %1 but %2 expected", prog,
2853 /* op must be string, result is number */
2854 propagate_types(b->left, c, ok, Tstr, 0);
2855 if (!type_compat(type, Tnum, 0))
2857 "error: Can only convert string to number, not %1",
2858 prog, type, 0, NULL);
2862 return propagate_types(b->right, c, ok, type, 0);
2864 ###### interp binode cases
2867 rv = interp_exec(b->left, &rvtype);
2868 right = interp_exec(b->right, &rtype);
2869 mpq_add(rv.num, rv.num, right.num);
2872 rv = interp_exec(b->left, &rvtype);
2873 right = interp_exec(b->right, &rtype);
2874 mpq_sub(rv.num, rv.num, right.num);
2877 rv = interp_exec(b->left, &rvtype);
2878 right = interp_exec(b->right, &rtype);
2879 mpq_mul(rv.num, rv.num, right.num);
2882 rv = interp_exec(b->left, &rvtype);
2883 right = interp_exec(b->right, &rtype);
2884 mpq_div(rv.num, rv.num, right.num);
2889 left = interp_exec(b->left, <ype);
2890 right = interp_exec(b->right, &rtype);
2891 mpz_init(l); mpz_init(r); mpz_init(rem);
2892 mpz_tdiv_q(l, mpq_numref(left.num), mpq_denref(left.num));
2893 mpz_tdiv_q(r, mpq_numref(right.num), mpq_denref(right.num));
2894 mpz_tdiv_r(rem, l, r);
2895 val_init(Tnum, &rv);
2896 mpq_set_z(rv.num, rem);
2897 mpz_clear(r); mpz_clear(l); mpz_clear(rem);
2902 rv = interp_exec(b->right, &rvtype);
2903 mpq_neg(rv.num, rv.num);
2906 rv = interp_exec(b->right, &rvtype);
2907 mpq_abs(rv.num, rv.num);
2910 rv = interp_exec(b->right, &rvtype);
2913 left = interp_exec(b->left, <ype);
2914 right = interp_exec(b->right, &rtype);
2916 rv.str = text_join(left.str, right.str);
2919 right = interp_exec(b->right, &rvtype);
2923 struct text tx = right.str;
2926 if (tx.txt[0] == '-') {
2931 if (number_parse(rv.num, tail, tx) == 0)
2934 mpq_neg(rv.num, rv.num);
2936 printf("Unsupported suffix: %.*s\n", tx.len, tx.txt);
2940 ###### value functions
2942 static struct text text_join(struct text a, struct text b)
2945 rv.len = a.len + b.len;
2946 rv.txt = malloc(rv.len);
2947 memcpy(rv.txt, a.txt, a.len);
2948 memcpy(rv.txt+a.len, b.txt, b.len);
2952 ### Blocks, Statements, and Statement lists.
2954 Now that we have expressions out of the way we need to turn to
2955 statements. There are simple statements and more complex statements.
2956 Simple statements do not contain (syntactic) newlines, complex statements do.
2958 Statements often come in sequences and we have corresponding simple
2959 statement lists and complex statement lists.
2960 The former comprise only simple statements separated by semicolons.
2961 The later comprise complex statements and simple statement lists. They are
2962 separated by newlines. Thus the semicolon is only used to separate
2963 simple statements on the one line. This may be overly restrictive,
2964 but I'm not sure I ever want a complex statement to share a line with
2967 Note that a simple statement list can still use multiple lines if
2968 subsequent lines are indented, so
2970 ###### Example: wrapped simple statement list
2975 is a single simple statement list. This might allow room for
2976 confusion, so I'm not set on it yet.
2978 A simple statement list needs no extra syntax. A complex statement
2979 list has two syntactic forms. It can be enclosed in braces (much like
2980 C blocks), or it can be introduced by an indent and continue until an
2981 unindented newline (much like Python blocks). With this extra syntax
2982 it is referred to as a block.
2984 Note that a block does not have to include any newlines if it only
2985 contains simple statements. So both of:
2987 if condition: a=b; d=f
2989 if condition { a=b; print f }
2993 In either case the list is constructed from a `binode` list with
2994 `Block` as the operator. When parsing the list it is most convenient
2995 to append to the end, so a list is a list and a statement. When using
2996 the list it is more convenient to consider a list to be a statement
2997 and a list. So we need a function to re-order a list.
2998 `reorder_bilist` serves this purpose.
3000 The only stand-alone statement we introduce at this stage is `pass`
3001 which does nothing and is represented as a `NULL` pointer in a `Block`
3002 list. Other stand-alone statements will follow once the infrastructure
3013 Block -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3014 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3015 | SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3016 | SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3017 | IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3019 OpenBlock -> OpenScope { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3020 | OpenScope { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3021 | OpenScope SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3022 | OpenScope SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3023 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3025 UseBlock -> { OpenScope IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3026 | { OpenScope SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3027 | IN OpenScope OptNL Statementlist OUT ${ $0 = $<Sl; }$
3029 ColonBlock -> { IN OptNL Statementlist OUT OptNL } ${ $0 = $<Sl; }$
3030 | { SimpleStatements } ${ $0 = reorder_bilist($<SS); }$
3031 | : SimpleStatements ; ${ $0 = reorder_bilist($<SS); }$
3032 | : SimpleStatements EOL ${ $0 = reorder_bilist($<SS); }$
3033 | : IN OptNL Statementlist OUT ${ $0 = $<Sl; }$
3035 Statementlist -> ComplexStatements ${ $0 = reorder_bilist($<CS); }$
3037 ComplexStatements -> ComplexStatements ComplexStatement ${
3047 | ComplexStatement ${
3059 ComplexStatement -> SimpleStatements Newlines ${
3060 $0 = reorder_bilist($<SS);
3062 | SimpleStatements ; Newlines ${
3063 $0 = reorder_bilist($<SS);
3065 ## ComplexStatement Grammar
3068 SimpleStatements -> SimpleStatements ; SimpleStatement ${
3074 | SimpleStatement ${
3082 SimpleStatement -> pass ${ $0 = NULL; }$
3083 | ERROR ${ tok_err(c, "Syntax error in statement", &$1); }$
3084 ## SimpleStatement Grammar
3086 ###### print binode cases
3090 if (b->left == NULL)
3093 print_exec(b->left, indent, bracket);
3096 print_exec(b->right, indent, bracket);
3099 // block, one per line
3100 if (b->left == NULL)
3101 do_indent(indent, "pass\n");
3103 print_exec(b->left, indent, bracket);
3105 print_exec(b->right, indent, bracket);
3109 ###### propagate binode cases
3112 /* If any statement returns something other than Tnone
3113 * or Tbool then all such must return same type.
3114 * As each statement may be Tnone or something else,
3115 * we must always pass NULL (unknown) down, otherwise an incorrect
3116 * error might occur. We never return Tnone unless it is
3121 for (e = b; e; e = cast(binode, e->right)) {
3122 t = propagate_types(e->left, c, ok, NULL, rules);
3123 if ((rules & Rboolok) && t == Tbool)
3125 if (t && t != Tnone && t != Tbool) {
3129 type_err(c, "error: expected %1%r, found %2",
3130 e->left, type, rules, t);
3136 ###### interp binode cases
3138 while (rvtype == Tnone &&
3141 rv = interp_exec(b->left, &rvtype);
3142 b = cast(binode, b->right);
3146 ### The Print statement
3148 `print` is a simple statement that takes a comma-separated list of
3149 expressions and prints the values separated by spaces and terminated
3150 by a newline. No control of formatting is possible.
3152 `print` faces the same list-ordering issue as blocks, and uses the
3158 ##### expr precedence
3161 ###### SimpleStatement Grammar
3163 | print ExpressionList ${
3164 $0 = reorder_bilist($<2);
3166 | print ExpressionList , ${
3171 $0 = reorder_bilist($0);
3182 ExpressionList -> ExpressionList , Expression ${
3195 ###### print binode cases
3198 do_indent(indent, "print");
3202 print_exec(b->left, -1, bracket);
3206 b = cast(binode, b->right);
3212 ###### propagate binode cases
3215 /* don't care but all must be consistent */
3216 propagate_types(b->left, c, ok, NULL, Rnolabel);
3217 propagate_types(b->right, c, ok, NULL, Rnolabel);
3220 ###### interp binode cases
3226 for ( ; b; b = cast(binode, b->right))
3230 left = interp_exec(b->left, <ype);
3231 print_value(ltype, &left);
3232 free_value(ltype, &left);
3243 ###### Assignment statement
3245 An assignment will assign a value to a variable, providing it hasn't
3246 been declared as a constant. The analysis phase ensures that the type
3247 will be correct so the interpreter just needs to perform the
3248 calculation. There is a form of assignment which declares a new
3249 variable as well as assigning a value. If a name is assigned before
3250 it is declared, and error will be raised as the name is created as
3251 `Tlabel` and it is illegal to assign to such names.
3257 ###### declare terminals
3260 ###### SimpleStatement Grammar
3261 | Variable = Expression ${
3267 | VariableDecl = Expression ${
3275 if ($1->var->where_set == NULL) {
3277 "Variable declared with no type or value: %v",
3287 ###### print binode cases
3290 do_indent(indent, "");
3291 print_exec(b->left, indent, bracket);
3293 print_exec(b->right, indent, bracket);
3300 struct variable *v = cast(var, b->left)->var;
3301 do_indent(indent, "");
3302 print_exec(b->left, indent, bracket);
3303 if (cast(var, b->left)->var->constant) {
3304 if (v->where_decl == v->where_set) {
3306 type_print(v->type, stdout);
3311 if (v->where_decl == v->where_set) {
3313 type_print(v->type, stdout);
3320 print_exec(b->right, indent, bracket);
3327 ###### propagate binode cases
3331 /* Both must match and not be labels,
3332 * Type must support 'dup',
3333 * For Assign, left must not be constant.
3336 t = propagate_types(b->left, c, ok, NULL,
3337 Rnolabel | (b->op == Assign ? Rnoconstant : 0));
3342 if (propagate_types(b->right, c, ok, t, 0) != t)
3343 if (b->left->type == Xvar)
3344 type_err(c, "info: variable '%v' was set as %1 here.",
3345 cast(var, b->left)->var->where_set, t, rules, NULL);
3347 t = propagate_types(b->right, c, ok, NULL, Rnolabel);
3349 propagate_types(b->left, c, ok, t,
3350 (b->op == Assign ? Rnoconstant : 0));
3352 if (t && t->dup == NULL)
3353 type_err(c, "error: cannot assign value of type %1", b, t, 0, NULL);
3358 ###### interp binode cases
3361 lleft = linterp_exec(b->left, <ype);
3362 right = interp_exec(b->right, &rtype);
3364 free_value(ltype, lleft);
3365 dup_value(ltype, &right, lleft);
3372 struct variable *v = cast(var, b->left)->var;
3375 free_value(v->type, v->val);
3378 right = interp_exec(b->right, &rtype);
3379 v->val = val_alloc(v->type, &right);
3382 v->val = val_alloc(v->type, NULL);
3387 ### The `use` statement
3389 The `use` statement is the last "simple" statement. It is needed when
3390 the condition in a conditional statement is a block. `use` works much
3391 like `return` in C, but only completes the `condition`, not the whole
3397 ###### expr precedence
3400 ###### SimpleStatement Grammar
3402 $0 = new_pos(binode, $1);
3405 if ($0->right->type == Xvar) {
3406 struct var *v = cast(var, $0->right);
3407 if (v->var->type == Tnone) {
3408 /* Convert this to a label */
3409 v->var->type = Tlabel;
3410 v->var->val = val_alloc(Tlabel, NULL);
3411 v->var->val->label = v->var->val;
3416 ###### print binode cases
3419 do_indent(indent, "use ");
3420 print_exec(b->right, -1, bracket);
3425 ###### propagate binode cases
3428 /* result matches value */
3429 return propagate_types(b->right, c, ok, type, 0);
3431 ###### interp binode cases
3434 rv = interp_exec(b->right, &rvtype);
3437 ### The Conditional Statement
3439 This is the biggy and currently the only complex statement. This
3440 subsumes `if`, `while`, `do/while`, `switch`, and some parts of `for`.
3441 It is comprised of a number of parts, all of which are optional though
3442 set combinations apply. Each part is (usually) a key word (`then` is
3443 sometimes optional) followed by either an expression or a code block,
3444 except the `casepart` which is a "key word and an expression" followed
3445 by a code block. The code-block option is valid for all parts and,
3446 where an expression is also allowed, the code block can use the `use`
3447 statement to report a value. If the code block does not report a value
3448 the effect is similar to reporting `True`.
3450 The `else` and `case` parts, as well as `then` when combined with
3451 `if`, can contain a `use` statement which will apply to some
3452 containing conditional statement. `for` parts, `do` parts and `then`
3453 parts used with `for` can never contain a `use`, except in some
3454 subordinate conditional statement.
3456 If there is a `forpart`, it is executed first, only once.
3457 If there is a `dopart`, then it is executed repeatedly providing
3458 always that the `condpart` or `cond`, if present, does not return a non-True
3459 value. `condpart` can fail to return any value if it simply executes
3460 to completion. This is treated the same as returning `True`.
3462 If there is a `thenpart` it will be executed whenever the `condpart`
3463 or `cond` returns True (or does not return any value), but this will happen
3464 *after* `dopart` (when present).
3466 If `elsepart` is present it will be executed at most once when the
3467 condition returns `False` or some value that isn't `True` and isn't
3468 matched by any `casepart`. If there are any `casepart`s, they will be
3469 executed when the condition returns a matching value.
3471 The particular sorts of values allowed in case parts has not yet been
3472 determined in the language design, so nothing is prohibited.
3474 The various blocks in this complex statement potentially provide scope
3475 for variables as described earlier. Each such block must include the
3476 "OpenScope" nonterminal before parsing the block, and must call
3477 `var_block_close()` when closing the block.
3479 The code following "`if`", "`switch`" and "`for`" does not get its own
3480 scope, but is in a scope covering the whole statement, so names
3481 declared there cannot be redeclared elsewhere. Similarly the
3482 condition following "`while`" is in a scope the covers the body
3483 ("`do`" part) of the loop, and which does not allow conditional scope
3484 extension. Code following "`then`" (both looping and non-looping),
3485 "`else`" and "`case`" each get their own local scope.
3487 The type requirements on the code block in a `whilepart` are quite
3488 unusal. It is allowed to return a value of some identifiable type, in
3489 which case the loop aborts and an appropriate `casepart` is run, or it
3490 can return a Boolean, in which case the loop either continues to the
3491 `dopart` (on `True`) or aborts and runs the `elsepart` (on `False`).
3492 This is different both from the `ifpart` code block which is expected to
3493 return a Boolean, or the `switchpart` code block which is expected to
3494 return the same type as the casepart values. The correct analysis of
3495 the type of the `whilepart` code block is the reason for the
3496 `Rboolok` flag which is passed to `propagate_types()`.
3498 The `cond_statement` cannot fit into a `binode` so a new `exec` is
3507 struct exec *action;
3508 struct casepart *next;
3510 struct cond_statement {
3512 struct exec *forpart, *condpart, *dopart, *thenpart, *elsepart;
3513 struct casepart *casepart;
3516 ###### ast functions
3518 static void free_casepart(struct casepart *cp)
3522 free_exec(cp->value);
3523 free_exec(cp->action);
3530 static void free_cond_statement(struct cond_statement *s)
3534 free_exec(s->forpart);
3535 free_exec(s->condpart);
3536 free_exec(s->dopart);
3537 free_exec(s->thenpart);
3538 free_exec(s->elsepart);
3539 free_casepart(s->casepart);
3543 ###### free exec cases
3544 case Xcond_statement: free_cond_statement(cast(cond_statement, e)); break;
3546 ###### ComplexStatement Grammar
3547 | CondStatement ${ $0 = $<1; }$
3549 ###### expr precedence
3550 $TERM for then while do
3557 // A CondStatement must end with EOL, as does CondSuffix and
3559 // ForPart, ThenPart, SwitchPart, CasePart are non-empty and
3560 // may or may not end with EOL
3561 // WhilePart and IfPart include an appropriate Suffix
3564 // Both ForPart and Whilepart open scopes, and CondSuffix only
3565 // closes one - so in the first branch here we have another to close.
3566 CondStatement -> ForPart OptNL ThenPart OptNL WhilePart CondSuffix ${
3569 $0->thenpart = $<TP;
3570 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3571 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3572 var_block_close(c, CloseSequential);
3574 | ForPart OptNL WhilePart CondSuffix ${
3577 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3578 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3579 var_block_close(c, CloseSequential);
3581 | WhilePart CondSuffix ${
3583 $0->condpart = $WP.condpart; $WP.condpart = NULL;
3584 $0->dopart = $WP.dopart; $WP.dopart = NULL;
3586 | SwitchPart OptNL CasePart CondSuffix ${
3588 $0->condpart = $<SP;
3589 $CP->next = $0->casepart;
3590 $0->casepart = $<CP;
3592 | SwitchPart : IN OptNL CasePart CondSuffix OUT Newlines ${
3594 $0->condpart = $<SP;
3595 $CP->next = $0->casepart;
3596 $0->casepart = $<CP;
3598 | IfPart IfSuffix ${
3600 $0->condpart = $IP.condpart; $IP.condpart = NULL;
3601 $0->thenpart = $IP.thenpart; $IP.thenpart = NULL;
3602 // This is where we close an "if" statement
3603 var_block_close(c, CloseSequential);
3606 CondSuffix -> IfSuffix ${
3608 // This is where we close scope of the whole
3609 // "for" or "while" statement
3610 var_block_close(c, CloseSequential);
3612 | Newlines CasePart CondSuffix ${
3614 $CP->next = $0->casepart;
3615 $0->casepart = $<CP;
3617 | CasePart CondSuffix ${
3619 $CP->next = $0->casepart;
3620 $0->casepart = $<CP;
3623 IfSuffix -> Newlines ${ $0 = new(cond_statement); }$
3624 | Newlines ElsePart ${ $0 = $<EP; }$
3625 | ElsePart ${$0 = $<EP; }$
3627 ElsePart -> else OpenBlock Newlines ${
3628 $0 = new(cond_statement);
3629 $0->elsepart = $<OB;
3630 var_block_close(c, CloseElse);
3632 | else OpenScope CondStatement ${
3633 $0 = new(cond_statement);
3634 $0->elsepart = $<CS;
3635 var_block_close(c, CloseElse);
3639 CasePart -> case Expression OpenScope ColonBlock ${
3640 $0 = calloc(1,sizeof(struct casepart));
3643 var_block_close(c, CloseParallel);
3647 // These scopes are closed in CondSuffix
3648 ForPart -> for OpenBlock ${
3652 ThenPart -> then OpenBlock ${
3654 var_block_close(c, CloseSequential);
3658 // This scope is closed in CondSuffix
3659 WhilePart -> while UseBlock OptNL do Block ${
3663 | while OpenScope Expression ColonBlock ${
3664 $0.condpart = $<Exp;
3668 IfPart -> if UseBlock OptNL then OpenBlock ClosePara ${
3672 | if OpenScope Expression OpenScope ColonBlock ClosePara ${
3676 | if OpenScope Expression OpenScope OptNL then Block ClosePara ${
3682 // This scope is closed in CondSuffix
3683 SwitchPart -> switch OpenScope Expression ${
3686 | switch UseBlock ${
3690 ###### print exec cases
3692 case Xcond_statement:
3694 struct cond_statement *cs = cast(cond_statement, e);
3695 struct casepart *cp;
3697 do_indent(indent, "for");
3698 if (bracket) printf(" {\n"); else printf("\n");
3699 print_exec(cs->forpart, indent+1, bracket);
3702 do_indent(indent, "} then {\n");
3704 do_indent(indent, "then\n");
3705 print_exec(cs->thenpart, indent+1, bracket);
3707 if (bracket) do_indent(indent, "}\n");
3711 if (cs->condpart && cs->condpart->type == Xbinode &&
3712 cast(binode, cs->condpart)->op == Block) {
3714 do_indent(indent, "while {\n");
3716 do_indent(indent, "while\n");
3717 print_exec(cs->condpart, indent+1, bracket);
3719 do_indent(indent, "} do {\n");
3721 do_indent(indent, "do\n");
3722 print_exec(cs->dopart, indent+1, bracket);
3724 do_indent(indent, "}\n");
3726 do_indent(indent, "while ");
3727 print_exec(cs->condpart, 0, bracket);
3732 print_exec(cs->dopart, indent+1, bracket);
3734 do_indent(indent, "}\n");
3739 do_indent(indent, "switch");
3741 do_indent(indent, "if");
3742 if (cs->condpart && cs->condpart->type == Xbinode &&
3743 cast(binode, cs->condpart)->op == Block) {
3748 print_exec(cs->condpart, indent+1, bracket);
3750 do_indent(indent, "}\n");
3752 do_indent(indent, "then:\n");
3753 print_exec(cs->thenpart, indent+1, bracket);
3757 print_exec(cs->condpart, 0, bracket);
3763 print_exec(cs->thenpart, indent+1, bracket);
3765 do_indent(indent, "}\n");
3770 for (cp = cs->casepart; cp; cp = cp->next) {
3771 do_indent(indent, "case ");
3772 print_exec(cp->value, -1, 0);
3777 print_exec(cp->action, indent+1, bracket);
3779 do_indent(indent, "}\n");
3782 do_indent(indent, "else");
3787 print_exec(cs->elsepart, indent+1, bracket);
3789 do_indent(indent, "}\n");
3794 ###### propagate exec cases
3795 case Xcond_statement:
3797 // forpart and dopart must return Tnone
3798 // thenpart must return Tnone if there is a dopart,
3799 // otherwise it is like elsepart.
3801 // be bool if there is no casepart
3802 // match casepart->values if there is a switchpart
3803 // either be bool or match casepart->value if there
3805 // elsepart and casepart->action must match the return type
3806 // expected of this statement.
3807 struct cond_statement *cs = cast(cond_statement, prog);
3808 struct casepart *cp;
3810 t = propagate_types(cs->forpart, c, ok, Tnone, 0);
3811 if (!type_compat(Tnone, t, 0))
3813 t = propagate_types(cs->dopart, c, ok, Tnone, 0);
3814 if (!type_compat(Tnone, t, 0))
3817 t = propagate_types(cs->thenpart, c, ok, Tnone, 0);
3818 if (!type_compat(Tnone, t, 0))
3821 if (cs->casepart == NULL)
3822 propagate_types(cs->condpart, c, ok, Tbool, 0);
3824 /* Condpart must match case values, with bool permitted */
3826 for (cp = cs->casepart;
3827 cp && !t; cp = cp->next)
3828 t = propagate_types(cp->value, c, ok, NULL, 0);
3829 if (!t && cs->condpart)
3830 t = propagate_types(cs->condpart, c, ok, NULL, Rboolok);
3831 // Now we have a type (I hope) push it down
3833 for (cp = cs->casepart; cp; cp = cp->next)
3834 propagate_types(cp->value, c, ok, t, 0);
3835 propagate_types(cs->condpart, c, ok, t, Rboolok);
3838 // (if)then, else, and case parts must return expected type.
3839 if (!cs->dopart && !type)
3840 type = propagate_types(cs->thenpart, c, ok, NULL, rules);
3842 type = propagate_types(cs->elsepart, c, ok, NULL, rules);
3843 for (cp = cs->casepart;
3846 type = propagate_types(cp->action, c, ok, NULL, rules);
3849 propagate_types(cs->thenpart, c, ok, type, rules);
3850 propagate_types(cs->elsepart, c, ok, type, rules);
3851 for (cp = cs->casepart; cp ; cp = cp->next)
3852 propagate_types(cp->action, c, ok, type, rules);
3858 ###### interp exec cases
3859 case Xcond_statement:
3861 struct value v, cnd;
3862 struct type *vtype, *cndtype;
3863 struct casepart *cp;
3864 struct cond_statement *c = cast(cond_statement, e);
3867 interp_exec(c->forpart, NULL);
3870 cnd = interp_exec(c->condpart, &cndtype);
3873 if (!(cndtype == Tnone ||
3874 (cndtype == Tbool && cnd.bool != 0)))
3876 // cnd is Tnone or Tbool, doesn't need to be freed
3878 interp_exec(c->dopart, NULL);
3881 rv = interp_exec(c->thenpart, &rvtype);
3882 if (rvtype != Tnone || !c->dopart)
3884 free_value(rvtype, &rv);
3887 } while (c->dopart);
3889 for (cp = c->casepart; cp; cp = cp->next) {
3890 v = interp_exec(cp->value, &vtype);
3891 if (value_cmp(cndtype, vtype, &v, &cnd) == 0) {
3892 free_value(vtype, &v);
3893 free_value(cndtype, &cnd);
3894 rv = interp_exec(cp->action, &rvtype);
3897 free_value(vtype, &v);
3899 free_value(cndtype, &cnd);
3901 rv = interp_exec(c->elsepart, &rvtype);
3908 ### Top level structure
3910 All the language elements so far can be used in various places. Now
3911 it is time to clarify what those places are.
3913 At the top level of a file there will be a number of declarations.
3914 Many of the things that can be declared haven't been described yet,
3915 such as functions, procedures, imports, and probably more.
3916 For now there are two sorts of things that can appear at the top
3917 level. They are predefined constants, `struct` types, and the main
3918 program. While the syntax will allow the main program to appear
3919 multiple times, that will trigger an error if it is actually attempted.
3921 The various declarations do not return anything. They store the
3922 various declarations in the parse context.
3924 ###### Parser: grammar
3927 Ocean -> OptNL DeclarationList
3929 ## declare terminals
3936 DeclarationList -> Declaration
3937 | DeclarationList Declaration
3939 Declaration -> ERROR Newlines ${
3941 "error: unhandled parse error", &$1);
3947 ## top level grammar
3949 ### The `const` section
3951 As well as being defined in with the code that uses them, constants
3952 can be declared at the top level. These have full-file scope, so they
3953 are always `InScope`. The value of a top level constant can be given
3954 as an expression, and this is evaluated immediately rather than in the
3955 later interpretation stage. Once we add functions to the language, we
3956 will need rules concern which, if any, can be used to define a top
3959 Constants are defined in a section that starts with the reserved word
3960 `const` and then has a block with a list of assignment statements.
3961 For syntactic consistency, these must use the double-colon syntax to
3962 make it clear that they are constants. Type can also be given: if
3963 not, the type will be determined during analysis, as with other
3966 As the types constants are inserted at the head of a list, printing
3967 them in the same order that they were read is not straight forward.
3968 We take a quadratic approach here and count the number of constants
3969 (variables of depth 0), then count down from there, each time
3970 searching through for the Nth constant for decreasing N.
3972 ###### top level grammar
3976 DeclareConstant -> const { IN OptNL ConstList OUT OptNL } Newlines
3977 | const { SimpleConstList } Newlines
3978 | const IN OptNL ConstList OUT Newlines
3979 | const SimpleConstList Newlines
3981 ConstList -> ConstList SimpleConstLine
3983 SimpleConstList -> SimpleConstList ; Const
3986 SimpleConstLine -> SimpleConstList Newlines
3987 | ERROR Newlines ${ tok_err(c, "Syntax error in constant", &$1); }$
3990 CType -> Type ${ $0 = $<1; }$
3993 Const -> IDENTIFIER :: CType = Expression ${ {
3997 v = var_decl(c, $1.txt);
3999 struct var *var = new_pos(var, $1);
4000 v->where_decl = var;
4005 v = var_ref(c, $1.txt);
4006 tok_err(c, "error: name already declared", &$1);
4007 type_err(c, "info: this is where '%v' was first declared",
4008 v->where_decl, NULL, 0, NULL);
4012 propagate_types($5, c, &ok, $3, 0);
4017 struct value res = interp_exec($5, &v->type);
4018 v->val = val_alloc(v->type, &res);
4022 ###### print const decls
4027 while (target != 0) {
4029 for (v = context.in_scope; v; v=v->in_scope)
4030 if (v->depth == 0) {
4041 printf(" %.*s :: ", v->name->name.len, v->name->name.txt);
4042 type_print(v->type, stdout);
4044 if (v->type == Tstr)
4046 print_value(v->type, v->val);
4047 if (v->type == Tstr)
4055 ### Finally the whole program.
4057 Somewhat reminiscent of Pascal a (current) Ocean program starts with
4058 the keyword "program" and a list of variable names which are assigned
4059 values from command line arguments. Following this is a `block` which
4060 is the code to execute. Unlike Pascal, constants and other
4061 declarations come *before* the program.
4063 As this is the top level, several things are handled a bit
4065 The whole program is not interpreted by `interp_exec` as that isn't
4066 passed the argument list which the program requires. Similarly type
4067 analysis is a bit more interesting at this level.
4072 ###### top level grammar
4074 DeclareProgram -> Program ${ {
4076 type_err(c, "Program defined a second time",
4085 Program -> program OpenScope Varlist ColonBlock Newlines ${
4088 $0->left = reorder_bilist($<Vl);
4090 var_block_close(c, CloseSequential);
4091 if (c->scope_stack && !c->parse_error) abort();
4094 Varlist -> Varlist ArgDecl ${
4103 ArgDecl -> IDENTIFIER ${ {
4104 struct variable *v = var_decl(c, $1.txt);
4111 ###### print binode cases
4113 do_indent(indent, "program");
4114 for (b2 = cast(binode, b->left); b2; b2 = cast(binode, b2->right)) {
4116 print_exec(b2->left, 0, 0);
4122 print_exec(b->right, indent+1, bracket);
4124 do_indent(indent, "}\n");
4127 ###### propagate binode cases
4128 case Program: abort(); // NOTEST
4130 ###### core functions
4132 static int analyse_prog(struct exec *prog, struct parse_context *c)
4134 struct binode *b = cast(binode, prog);
4141 propagate_types(b->right, c, &ok, Tnone, 0);
4146 for (b = cast(binode, b->left); b; b = cast(binode, b->right)) {
4147 struct var *v = cast(var, b->left);
4148 if (!v->var->type) {
4149 v->var->where_set = b;
4150 v->var->type = Tstr;
4154 b = cast(binode, prog);
4157 propagate_types(b->right, c, &ok, Tnone, 0);
4162 /* Make sure everything is still consistent */
4163 propagate_types(b->right, c, &ok, Tnone, 0);
4167 static void interp_prog(struct exec *prog, char **argv)
4169 struct binode *p = cast(binode, prog);
4176 al = cast(binode, p->left);
4178 struct var *v = cast(var, al->left);
4179 struct value *vl = v->var->val;
4181 if (argv[0] == NULL) {
4182 printf("Not enough args\n");
4185 al = cast(binode, al->right);
4187 free_value(v->var->type, vl);
4189 vl = val_alloc(v->var->type, NULL);
4192 free_value(v->var->type, vl);
4193 vl->str.len = strlen(argv[0]);
4194 vl->str.txt = malloc(vl->str.len);
4195 memcpy(vl->str.txt, argv[0], vl->str.len);
4198 v = interp_exec(p->right, &vtype);
4199 free_value(vtype, &v);
4202 ###### interp binode cases
4203 case Program: abort(); // NOTEST
4205 ## And now to test it out.
4207 Having a language requires having a "hello world" program. I'll
4208 provide a little more than that: a program that prints "Hello world"
4209 finds the GCD of two numbers, prints the first few elements of
4210 Fibonacci, performs a binary search for a number, and a few other
4211 things which will likely grow as the languages grows.
4213 ###### File: oceani.mk
4216 @echo "===== DEMO ====="
4217 ./oceani --section "demo: hello" oceani.mdc 55 33
4223 four ::= 2 + 2 ; five ::= 10/2
4224 const pie ::= "I like Pie";
4225 cake ::= "The cake is"
4234 print "Hello World, what lovely oceans you have!"
4235 print "Are there", five, "?"
4236 print pi, pie, "but", cake
4238 A := $Astr; B := $Bstr
4240 /* When a variable is defined in both branches of an 'if',
4241 * and used afterwards, the variables are merged.
4247 print "Is", A, "bigger than", B,"? ", bigger
4248 /* If a variable is not used after the 'if', no
4249 * merge happens, so types can be different
4252 double:string = "yes"
4253 print A, "is more than twice", B, "?", double
4256 print "double", B, "is", double
4261 if a > 0 and then b > 0:
4267 print "GCD of", A, "and", B,"is", a
4269 print a, "is not positive, cannot calculate GCD"
4271 print b, "is not positive, cannot calculate GCD"
4276 print "Fibonacci:", f1,f2,
4277 then togo = togo - 1
4285 /* Binary search... */
4290 mid := (lo + hi) / 2
4302 print "Yay, I found", target
4304 print "Closest I found was", mid
4309 // "middle square" PRNG. Not particularly good, but one my
4310 // Dad taught me - the first one I ever heard of.
4311 for i:=1; then i = i + 1; while i < size:
4312 n := list[i-1] * list[i-1]
4313 list[i] = (n / 100) % 10 000
4315 print "Before sort:",
4316 for i:=0; then i = i + 1; while i < size:
4320 for i := 1; then i=i+1; while i < size:
4321 for j:=i-1; then j=j-1; while j >= 0:
4322 if list[j] > list[j+1]:
4326 print " After sort:",
4327 for i:=0; then i = i + 1; while i < size:
4331 if 1 == 2 then print "yes"; else print "no"
4335 bob.alive = (bob.name == "Hello")
4336 print "bob", "is" if bob.alive else "isn't", "alive"