implement: unit identity -- dotted paths, name rules, source path

Units start here, with the part that needs no import graph: what a unit is
called and where it must live.

The parser only ever read a single identifier after `unit` and `import`, so
`unit game.main;` and `import std.io;` were syntax errors -- which is why the
dotted fixtures failed at the semicolon. It now reads a dotted path and stores
it canonically, dots included, since that spelling is the unit's identity
everywhere else. `import a.b as c;` parses too, with the alias on the node.

resolve.c is the new pass between parsing and checking, for the questions that
span files. It carries SPEC 8.1 so far: each path segment is ASCII lowercase,
starts with a letter, continues with letters, digits or underscore, and is at
most eight characters; and the dotted path must match the source path it was
read from, so game.world.map has to come from game/world/map.fe. The source
side is folded to lowercase before comparing, because a case-insensitive host
must not let two spellings become two units.

That rule then applied to the fixtures, which were not obeying it: 57 declared
a unit name unrelated to their file, left over from the milestone directories,
and eight had names too long to be legal. Both are now aligned -- the rule is
worth having only if the tree follows it.

units: badupper, badlong and unitbad pass. 138 -> 146 of 188. The rest of
units/ needs the import graph, which is the next piece: resolution, cycles,
bindings and visibility.

fec/src/driver.c

@@ -1,5 +1,6 @@
 #include "parser.h"
 #include "check.h"
+#include "resolve.h"
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -80,6 +81,15 @@ int main(int argc, char **argv)
         free(src);
         return d.errors?1:0;
     }
+    /* Unit identity before semantic analysis: a unit that is not named
+       correctly, or does not sit where its name says, cannot be resolved from
+       another unit either. */
+    fe_resolve_unit_identity(&ast,&d,file);
+    if(d.errors){
+        fe_ast_destroy(&ast);
+        free(src);
+        return 1;
+    }
     fe_check_init(&check,&ast,&d,pointer_bits,no_checks);
     if(!fe_check_program(&check)){
         fe_ast_destroy(&ast);

fec/src/parser.c

@@ -302,12 +302,46 @@ static FeNode *statement(FeParser *p)
     e=expr(p,0); if(is(p,FE_TOK_EQ)||is(p,FE_TOK_PLUS_EQ)||is(p,FE_TOK_MINUS_EQ)||is(p,FE_TOK_STAR_EQ)||is(p,FE_TOK_SLASH_EQ)||is(p,FE_TOK_PERCENT_EQ)||is(p,FE_TOK_AND_EQ)||is(p,FE_TOK_OR_EQ)||is(p,FE_TOK_XOR_EQ)||is(p,FE_TOK_SHL_EQ)||is(p,FE_TOK_SHR_EQ)){n=toknode(p,FE_N_ASSIGN,p->current);n->a=e;next(p);n->b=expr(p,0);}else{n=toknode(p,FE_N_EXPR_STMT,t);n->a=e;}want(p,FE_TOK_SEMI,"expected ';' after statement");return n;
 }
 
+/* A unit path is dotted: `game.world.map`. It is stored canonically, dots and
+   all, because that spelling is the unit's identity everywhere else. */
+static char *unit_path(FeParser *p)
+{
+    char buf[256];
+    unsigned long len=0;
+    if(!is_name(p)) return 0;
+    for(;;) {
+        unsigned long n=p->current.length;
+        if(len && len+1<sizeof buf) buf[len++]='.';
+        if(len+n>=sizeof buf){error(p,"unit path is too long");return 0;}
+        memcpy(buf+len,p->current.begin,n);
+        len+=n;
+        next(p);
+        if(!eat(p,FE_TOK_DOT)) break;
+        if(!is_name(p)){error(p,"expected a name after '.' in unit path");return 0;}
+    }
+    return fe_arena_strdup(&p->ast->arena,buf,len);
+}
+
 FeNode *fe_parse_unit(FeParser *p)
 {
-    FeToken t=p->current, name; FeNode *root;
+    FeToken t=p->current; FeNode *root; char *path;
     if(!eat(p,FE_TOK_UNIT)){error(p,"source must start with 'unit'");return fe_node(p->ast,FE_N_ERROR_NODE,t.loc,"unit",4);}
-    root=toknode(p,FE_N_UNIT,t);if(is_name(p)){name=p->current;root->text=fe_arena_strdup(&p->ast->arena,name.begin,name.length);next(p);}else error(p,"expected unit name");want(p,FE_TOK_SEMI,"expected ';' after unit name");
-    while(eat(p,FE_TOK_IMPORT)){FeToken it=p->previous;FeNode *i=toknode(p,FE_N_IMPORT,it);if(is_name(p)){next(p);i->text=fe_arena_strdup(&p->ast->arena,p->previous.begin,p->previous.length);}else error(p,"expected import name");want(p,FE_TOK_SEMI,"expected ';' after import");fe_node_add(root,i);}
+    root=toknode(p,FE_N_UNIT,t);
+    path=unit_path(p);
+    if(path) root->text=path; else error(p,"expected unit name");
+    want(p,FE_TOK_SEMI,"expected ';' after unit name");
+    while(eat(p,FE_TOK_IMPORT)){
+        FeToken it=p->previous;FeNode *i=toknode(p,FE_N_IMPORT,it);
+        path=unit_path(p);
+        if(path) i->text=path; else error(p,"expected import name");
+        /* `as` renames the binding; without it the binding is the last segment. */
+        if(eat(p,FE_TOK_AS)) {
+            if(is_name(p)){i->aux_text=fe_arena_strdup(&p->ast->arena,p->current.begin,p->current.length);next(p);}
+            else error(p,"expected an alias name after 'as'");
+        }
+        want(p,FE_TOK_SEMI,"expected ';' after import");
+        fe_node_add(root,i);
+    }
     while(!is(p,FE_TOK_EOF)){FeNode *d=decl(p);if(d)fe_node_add(root,d);}
     return root;
 }

fec/src/resolve.c

@@ -0,0 +1,86 @@
+#include "resolve.h"
+
+#include <string.h>
+
+static int segment_ok(const char *s, unsigned long n, const char **why)
+{
+    unsigned long i;
+    if (!n) { *why = "unit path segment is empty"; return 0; }
+    if (n > FE_UNIT_SEGMENT_MAX) {
+        *why = "unit path segment is longer than eight characters";
+        return 0;
+    }
+    if (s[0] < 'a' || s[0] > 'z') {
+        *why = "unit path segment must start with a lowercase letter";
+        return 0;
+    }
+    for (i = 1; i < n; ++i) {
+        char c = s[i];
+        if ((c >= 'a' && c <= 'z') || (c >= '0' && c <= '9') || c == '_') continue;
+        *why = "unit path segment may only contain lowercase letters, digits and '_'";
+        return 0;
+    }
+    return 1;
+}
+
+/* Compare a dotted unit path against the source path it was read from.
+   `game.world.map` matches `.../game/world/map.fe` and nothing else. Only the
+   trailing segments are compared, since the leading part is the import root. */
+static int path_matches(const char *unit, const char *source)
+{
+    unsigned long ulen = strlen(unit), slen = strlen(source);
+    unsigned long u, s;
+    if (slen < 3 || strcmp(source + slen - 3, ".fe") != 0) return 0;
+    slen -= 3;
+    u = ulen;
+    s = slen;
+    while (u > 0) {
+        char uc, sc;
+        --u;
+        if (s == 0) return 0;
+        --s;
+        uc = unit[u];
+        sc = source[s];
+        if (uc == '.') {
+            if (sc != '/' && sc != '\\') return 0;
+            continue;
+        }
+        /* Host filesystems may be case-insensitive; the unit name is the
+           authority and is lowercase by 8.1, so fold the path side down. */
+        if (sc >= 'A' && sc <= 'Z') sc = (char)(sc - 'A' + 'a');
+        if (uc != sc) return 0;
+    }
+    /* What remains of the source path is the import root, and must end there. */
+    return s == 0 || source[s - 1] == '/' || source[s - 1] == '\\';
+}
+
+int fe_resolve_unit_identity(FeAst *ast, FeDiags *diags, const char *source_path)
+{
+    FeNode *root = ast ? ast->root : 0;
+    const char *name, *why;
+    const char *seg;
+    unsigned long i, len;
+    int ok = 1;
+
+    if (!root || root->kind != FE_N_UNIT || !root->text) return 0;
+    name = root->text;
+    len = strlen(name);
+
+    seg = name;
+    for (i = 0; i <= len; ++i) {
+        if (i != len && name[i] != '.') continue;
+        if (!segment_ok(seg, (unsigned long)(name + i - seg), &why)) {
+            fe_diag_error(diags, root->loc, why);
+            ok = 0;
+        }
+        seg = name + i + 1;
+    }
+
+    if (ok && source_path && !path_matches(name, source_path)) {
+        fe_diag_errorf(diags, root->loc,
+                       "unit %s must be declared in a source file matching its path",
+                       name);
+        ok = 0;
+    }
+    return ok;
+}

fec/src/resolve.h

@@ -0,0 +1,24 @@
+#ifndef FE_RESOLVE_H
+#define FE_RESOLVE_H
+
+#include "ast.h"
+#include "diag.h"
+
+/* Unit-level resolution: identity, import bindings, and the unit graph.
+   This runs between parsing and semantic checking. It answers questions that
+   need more than one file -- what a unit is called, what it imports, and
+   whether those imports exist and terminate -- so that check.c can keep
+   looking at one function at a time. */
+
+/* SPEC 8.1: each segment is ASCII lowercase, starts with a letter, continues
+   with letters, digits or '_', and is at most eight characters. The limit is
+   what makes a unit path map to a FAT/DOS 8.3 source path unambiguously. */
+#define FE_UNIT_SEGMENT_MAX 8
+
+/* Validate the `unit` declaration against SPEC 8.1, and against the file it was
+   read from: the path must match the dotted name, so `game.world.map` has to
+   come from `game/world/map.fe`. `source_path` may be null to skip that half.
+   Returns non-zero when the unit is well formed. */
+int fe_resolve_unit_identity(FeAst *ast, FeDiags *diags, const char *source_path);
+
+#endif

fec/tests/format/bad_ari.fe

@@ -1,4 +1,4 @@
-unit m4_bad_arity;
+unit bad_ari;
 
 fn main() -> i32 {
     @print("{} {}", 1);

fec/tests/format/bad_bufw.fe

@@ -1,4 +1,4 @@
-unit m4_bad_buffer_writer;
+unit bad_bufw;
 
 fn main() -> void {
     var raw: [4]u8 = [0, 0, 0, 0];

fec/tests/format/bad_cls.fe

@@ -1,4 +1,4 @@
-unit m4_bad_cls;
+unit bad_cls;
 
 fn main() -> i32 {
     @print("}", 1);

fec/tests/format/bad_many.fe

@@ -1,4 +1,4 @@
-unit m4_bad_many;
+unit bad_many;
 
 fn main() -> i32 {
     @print("{}", 1, 2);

fec/tests/format/bad_open.fe

@@ -1,4 +1,4 @@
-unit m4_bad_open;
+unit bad_open;
 
 fn main() -> i32 {
     @print("{", 1);

fec/tests/format/bad_run.fe

@@ -1,4 +1,4 @@
-unit m4_bad_runtime;
+unit bad_run;
 
 fn main() -> i32 {
     var fmt: str = "{}";

fec/tests/format/bad_try.fe

@@ -1,4 +1,4 @@
-unit m4_bad_try;
+unit bad_try;
 
 fn main() -> i32 {
     try @print("nope");

fec/tests/format/bad_type.fe

@@ -1,4 +1,4 @@
-unit m4_bad_type;
+unit bad_type;
 
 struct Point { x: i32, }
 

fec/tests/format/bad_verb.fe

@@ -1,4 +1,4 @@
-unit m4_bad_verb;
+unit bad_verb;
 
 fn main() -> i32 {
     @print("{q}", 1);

fec/tests/format/bad_writ.fe

@@ -1,4 +1,4 @@
-unit m4_bad_writer;
+unit bad_writ;
 
 fn main() -> i32 {
     var x: i32 = 0;

fec/tests/format/ok_forma.fe

@@ -1,4 +1,4 @@
-unit m4_format;
+unit ok_forma;
 
 const FMT: str = "n={} hex={x} c={c} s={s} b={b} {{ok}}\n";
 

fec/tests/format/ok_prop.fe

@@ -1,4 +1,4 @@
-unit m4_prop;
+unit ok_prop;
 
 pub fn propagate(w: io.Writer) -> !void {
     try @fprint(w, "a{}b", 1);

fec/tests/format/ok_try_f.fe

@@ -1,4 +1,4 @@
-unit m4_try_fprint;
+unit ok_try_f;
 
 fn main() -> !void {
     var raw: [4]u8 = [0, 0, 0, 0];

fec/tests/own/bad_clos.fe

@@ -1,4 +1,4 @@
-unit m5_bad_consuming_close;
+unit bad_clos;
 
 struct FileLike {
     handle: i32,

fec/tests/own/bad_cond.fe

@@ -1,4 +1,4 @@
-unit m5_bad_conditional;
+unit bad_cond;
 
 fn take(p: ^i32) -> void { mem.destroy(p); }
 

fec/tests/own/bad_dbl.fe

@@ -1,4 +1,4 @@
-unit m5_bad_double;
+unit bad_dbl;
 
 fn bad(p: ^i32) -> void {
     mem.destroy(p);

fec/tests/own/bad_dest.fe

@@ -1,4 +1,4 @@
-unit m5_bad_destroy;
+unit bad_dest;
 
 fn bad(x: i32) -> void {
     mem.destroy(x);

fec/tests/own/bad_drop.fe

@@ -1,4 +1,4 @@
-unit m5_bad_drop;
+unit bad_drop;
 
 struct Box {
     value: i32,

fec/tests/own/bad_loop.fe

@@ -1,4 +1,4 @@
-unit m5_bad_loop_move;
+unit bad_loop;
 
 fn take(p: ^i32) -> void { mem.destroy(p); }
 

fec/tests/own/bad_move.fe

@@ -1,4 +1,4 @@
-unit m5_bad_move;
+unit bad_move;
 
 fn take(p: ^i32) -> void { mem.destroy(p); }
 

fec/tests/own/bad_proj.fe

@@ -1,4 +1,4 @@
-unit m5_bad_projection_move;
+unit bad_proj;
 
 struct Holder { p: ^i32 }
 fn take(p: ^i32) -> void { mem.destroy(p); }

fec/tests/own/ok_defer.fe

@@ -1,4 +1,4 @@
-unit m5_defer;
+unit ok_defer;
 
 pub fn cleanup(p: ^i32) -> void {
     defer { mem.destroy(p); }

fec/tests/own/ok_owned.fe

@@ -1,4 +1,4 @@
-unit m5_owned;
+unit ok_owned;
 
 fn main() -> !void {
     var p: ^i32 = try mem.create(0);

fec/tests/parse/keybuilt.fe

@@ -1,4 +1,4 @@
-unit keywords_and_builtins;
+unit keybuilt;
 
 pub fn demo() {
     let a = true and not false;

fec/tests/parse/logical.fe

@@ -1,3 +1,3 @@
 // ERROR:logical operator
-unit old_logic;
+unit logical;
 fn main() { let x = true && false; }

fec/tests/parse/misssemi.fe

@@ -1,3 +1,3 @@
 // ERROR:expected ';'
-unit broken;
+unit misssemi;
 fn main() { let x: i32 = 1 }

fec/tests/parse/unclcomm.fe

@@ -1,3 +1,3 @@
 // ERROR:unterminated block comment
-unit broken;
+unit unclcomm;
 /* no ending delimiter

fec/tests/parse/v012form.fe

@@ -1,4 +1,4 @@
-unit v012_forms;
+unit v012form;
 
 shared atomic var ticks: u16 = 0;
 packed struct Packet {

fec/tests/types/bad_ari.fe

@@ -1,4 +1,4 @@
-unit bad_arity;
+unit bad_ari;
 
 fn add(a: i32, b: i32) -> i32 {
     return a + b;

fec/tests/types/bad_asgn.fe

@@ -1,4 +1,4 @@
-unit bad_assign;
+unit bad_asgn;
 
 fn main() -> i32 {
     let value: i32 = 1;

fec/tests/types/bad_cond.fe

@@ -1,4 +1,4 @@
-unit bad_condition;
+unit bad_cond;
 
 fn main() -> i32 {
     if 1 { return 0; }

fec/tests/types/bad_mlet.fe

@@ -1,4 +1,4 @@
-unit m3_bad_mut_let;
+unit bad_mlet;
 
 fn bad() -> void {
     var raw: [2]u8 = [1, 2];

fec/tests/types/bad_ret.fe

@@ -1,4 +1,4 @@
-unit bad_return;
+unit bad_ret;
 
 fn main() -> i32 {
     return true;

fec/tests/types/bad_shwr.fe

@@ -1,4 +1,4 @@
-unit m3_bad_shared_write;
+unit bad_shwr;
 
 fn bad(s: []u8) -> void {
     s[0] = 1;

fec/tests/types/bad_type.fe

@@ -1,4 +1,4 @@
-unit bad_types;
+unit bad_type;
 
 fn add(a: i32, b: i32) -> i32 {
     return a + b;

fec/tests/types/bad_unit.fe

@@ -1,4 +1,4 @@
-unit bad_uninit;
+unit bad_unit;
 
 fn main() -> i32 {
     var value: i32;

fec/tests/types/bad_unk.fe

@@ -1,4 +1,4 @@
-unit bad_unknown;
+unit bad_unk;
 
 fn main() -> i32 {
     return missing_name;

fec/tests/types/badarr.fe

@@ -1,4 +1,4 @@
-unit fail_m3_array;
+unit badarr;
 fn main() -> i32 {
     let a: [2]i32 = [1, true, 3];
     return a[0];

fec/tests/types/badchar.fe

@@ -1,4 +1,4 @@
-unit fail_m3_char;
+unit badchar;
 
 fn main() -> i32 {
     let u: u8 = 'A';

fec/tests/types/badcycle.fe

@@ -1,4 +1,4 @@
-unit fail_m3_cycle;
+unit badcycle;
 
 struct A { b: B, }
 struct B { a: A, }

fec/tests/types/badfield.fe

@@ -1,4 +1,4 @@
-unit fail_m3_let_field;
+unit badfield;
 
 struct Point { x: i32, y: i32, }
 fn main() -> i32 {

fec/tests/types/badfld.fe

@@ -1,4 +1,4 @@
-unit fail_m3_fields;
+unit badfld;
 struct Point { x: i32, y: i32, }
 fn main() -> i32 {
     let p: Point = Point{ x: 1 };

fec/tests/types/badindex.fe

@@ -1,4 +1,4 @@
-unit fail_m3_let_index;
+unit badindex;
 
 fn main() -> i32 {
     let a: [2]i32 = [1, 2];

fec/tests/types/badmat.fe

@@ -1,4 +1,4 @@
-unit fail_m3_match;
+unit badmat;
 enum Shape { Empty, Circle(i32), }
 fn main() -> i32 {
     match Shape.Empty { Empty => 0; }

fec/tests/types/badstr.fe

@@ -1,4 +1,4 @@
-unit fail_m3_str;
+unit badstr;
 fn main() -> i32 {
     var text: str = "abc";
     text[0] = 'z';

fec/tests/types/ok_arra1.fe

@@ -1,4 +1,4 @@
-unit m3_arrayctx;
+unit ok_arra1;
 
 fn main() -> i32 {
     let bytes: [3]u8 = [1, 2, 3];

fec/tests/types/ok_array.fe

@@ -1,4 +1,4 @@
-unit m3_array;
+unit ok_array;
 
 fn main() -> i32 {
     let a: [3]i32 = [1, 2, 3];

fec/tests/types/ok_castw.fe

@@ -1,4 +1,4 @@
-unit cast_while;
+unit ok_castw;
 
 pub fn main() -> i32 {
     var x: i16 = 0;

fec/tests/types/ok_char.fe

@@ -1,4 +1,4 @@
-unit m3_char;
+unit ok_char;
 
 fn main() -> i32 {
     let c: char = '\u0041';

fec/tests/types/ok_enum.fe

@@ -1,4 +1,4 @@
-unit m3_enum;
+unit ok_enum;
 
 enum Shape { Empty, Circle(i32), Rect { w: i32, h: i32, }, }
 

fec/tests/types/ok_for.fe

@@ -1,4 +1,4 @@
-unit m3_for;
+unit ok_for;
 
 fn main() -> i32 {
     var total: i32 = 0;

fec/tests/types/ok_hello.fe

@@ -1,4 +1,4 @@
-unit hello;
+unit ok_hello;
 
 fn add(a: i32, b: i32) -> i32 {
     return a + b;

fec/tests/types/ok_mutab.fe

@@ -1,4 +1,4 @@
-unit m3_mutable;
+unit ok_mutab;
 
 fn takes_shared(s: []u8) -> u8 { return s[0]; }
 

fec/tests/types/ok_neste.fe

@@ -1,4 +1,4 @@
-unit m3_nested;
+unit ok_neste;
 
 struct Outer { inner: Inner, }
 struct Inner { value: i32, }

fec/tests/types/ok_scope.fe

@@ -1,4 +1,4 @@
-unit scopes;
+unit ok_scope;
 
 fn register(switch: i32) -> i32 {
     let auto: i32 = switch;

fec/tests/types/ok_str.fe

@@ -1,4 +1,4 @@
-unit m3_str;
+unit ok_str;
 
 fn main() -> i32 {
     let text: str = "abc";

fec/tests/types/ok_struc.fe

@@ -1,4 +1,4 @@
-unit m3_struct;
+unit ok_struc;
 
 struct Point { x: i32, y: i32, }
 packed struct PackedPoint { x: u8, y: i32, }