Merge pull request #1344 from ianthehenry/peg-sub-special

Add a new (sub) PEG special

src/core/peg.c

@@ -39,6 +39,10 @@
 typedef struct {
     const uint8_t *text_start;
     const uint8_t *text_end;
+    /* text_end will be restricted in a (sub) rule, but
+       outer_text_end will always contain the real end of
+       input, which we need to generate a line mapping */
+    const uint8_t *outer_text_end;
     const uint32_t *bytecode;
     const Janet *constants;
     JanetArray *captures;
@@ -114,12 +118,12 @@ static LineCol get_linecol_from_position(PegState *s, int32_t position) {
     /* Generate if not made yet */
     if (s->linemaplen < 0) {
         int32_t newline_count = 0;
-        for (const uint8_t *c = s->text_start; c < s->text_end; c++) {
+        for (const uint8_t *c = s->text_start; c < s->outer_text_end; c++) {
             if (*c == '\n') newline_count++;
         }
         int32_t *mem = janet_smalloc(sizeof(int32_t) * newline_count);
         size_t index = 0;
-        for (const uint8_t *c = s->text_start; c < s->text_end; c++) {
+        for (const uint8_t *c = s->text_start; c < s->outer_text_end; c++) {
             if (*c == '\n') mem[index++] = (int32_t)(c - s->text_start);
         }
         s->linemaplen = newline_count;
@@ -179,7 +183,7 @@ static const uint8_t *peg_rule(
     const uint32_t *rule,
     const uint8_t *text) {
 tail:
-    switch (*rule & 0x1F) {
+    switch (*rule) {
         default:
             janet_panic("unexpected opcode");
             return NULL;
@@ -482,6 +486,30 @@ tail:
             return result;
         }
 
+        case RULE_SUB: {
+            const uint8_t *text_start = text;
+            const uint32_t *rule_window = s->bytecode + rule[1];
+            const uint32_t *rule_subpattern = s->bytecode + rule[2];
+            down1(s);
+            const uint8_t *window_end = peg_rule(s, rule_window, text);
+            up1(s);
+            if (!window_end) {
+                return NULL;
+            }
+            const uint8_t *saved_end = s->text_end;
+            s->text_end = window_end;
+            down1(s);
+            const uint8_t *next_text = peg_rule(s, rule_subpattern, text_start);
+            up1(s);
+            s->text_end = saved_end;
+
+            if (!next_text) {
+                return NULL;
+            }
+
+            return window_end;
+        }
+
         case RULE_REPLACE:
         case RULE_MATCHTIME: {
             uint32_t tag = rule[3];
@@ -1107,6 +1135,14 @@ static void spec_matchtime(Builder *b, int32_t argc, const Janet *argv) {
     emit_3(r, RULE_MATCHTIME, subrule, cindex, tag);
 }
 
+static void spec_sub(Builder *b, int32_t argc, const Janet *argv) {
+    peg_fixarity(b, argc, 2);
+    Reserve r = reserve(b, 3);
+    uint32_t subrule1 = peg_compile1(b, argv[0]);
+    uint32_t subrule2 = peg_compile1(b, argv[1]);
+    emit_2(r, RULE_SUB, subrule1, subrule2);
+}
+
 #ifdef JANET_INT_TYPES
 #define JANET_MAX_READINT_WIDTH 8
 #else
@@ -1190,6 +1226,7 @@ static const SpecialPair peg_specials[] = {
     {"sequence", spec_sequence},
     {"set", spec_set},
     {"some", spec_some},
+    {"sub", spec_sub},
     {"thru", spec_thru},
     {"to", spec_to},
     {"uint", spec_uint_le},
@@ -1431,7 +1468,7 @@ static void *peg_unmarshal(JanetMarshalContext *ctx) {
         uint32_t instr = bytecode[i];
         uint32_t *rule = bytecode + i;
         op_flags[i] |= 0x02;
-        switch (instr & 0x1F) {
+        switch (instr) {
             case RULE_LITERAL:
                 i += 2 + ((rule[1] + 3) >> 2);
                 break;
@@ -1524,6 +1561,14 @@ static void *peg_unmarshal(JanetMarshalContext *ctx) {
                 op_flags[rule[1]] |= 0x01;
                 i += 4;
                 break;
+            case RULE_SUB:
+                /* [rule, rule] */
+                if (rule[1] >= blen) goto bad;
+                if (rule[2] >= blen) goto bad;
+                op_flags[rule[1]] |= 0x01;
+                op_flags[rule[2]] |= 0x01;
+                i += 3;
+                break;
             case RULE_ERROR:
             case RULE_DROP:
             case RULE_NOT:
@@ -1677,6 +1722,7 @@ static PegCall peg_cfun_init(int32_t argc, Janet *argv, int get_replace) {
     ret.s.mode = PEG_MODE_NORMAL;
     ret.s.text_start = ret.bytes.bytes;
     ret.s.text_end = ret.bytes.bytes + ret.bytes.len;
+    ret.s.outer_text_end = ret.s.text_end;
     ret.s.depth = JANET_RECURSION_GUARD;
     ret.s.captures = janet_array(0);
     ret.s.tagged_captures = janet_array(0);

src/include/janet.h

@@ -2140,7 +2140,8 @@ typedef enum {
     RULE_LINE,         /* [tag] */
     RULE_COLUMN,       /* [tag] */
     RULE_UNREF,        /* [rule, tag] */
-    RULE_CAPTURE_NUM   /* [rule, tag] */
+    RULE_CAPTURE_NUM,  /* [rule, tag] */
+    RULE_SUB           /* [rule, rule] */
 } JanetPegOpcod;
 
 typedef struct {

test/suite-peg.janet

@@ -263,6 +263,8 @@
 (marshpeg '(if-not "abcdf" 123))
 (marshpeg ~(cmt "abcdf" ,identity))
 (marshpeg '(group "abc"))
+(marshpeg '(sub "abcdf" "abc"))
+(marshpeg '(* (sub 1 1)))
 
 # Peg swallowing errors
 # 159651117
@@ -660,5 +662,53 @@
   (peg/match '(if (not (* (constant 7) "a")) "hello") "hello")
   @[]) "peg if not")
 
+(defn test [name peg input expected]
+  (assert (deep= (peg/match peg input) expected) name))
+
+(test "sub: matches the same input twice"
+  ~(sub "abcd" "abc")
+  "abcdef"
+  @[])
+
+(test "sub: second pattern cannot match more than the first pattern"
+  ~(sub "abcd" "abcde")
+  "abcdef"
+  nil)
+
+(test "sub: fails if first pattern fails"
+  ~(sub "x" "abc")
+  "abcdef"
+  nil)
+
+(test "sub: fails if second pattern fails"
+  ~(sub "abc" "x")
+  "abcdef"
+  nil)
+
+(test "sub: keeps captures from both patterns"
+  ~(sub '"abcd" '"abc")
+  "abcdef"
+  @["abcd" "abc"])
+
+(test "sub: second pattern can reference captures from first"
+  ~(* (constant 5 :tag) (sub (capture "abc" :tag) (backref :tag)))
+  "abcdef"
+  @[5 "abc" "abc"])
+
+(test "sub: second pattern can't see past what the first pattern matches"
+  ~(sub "abc" (* "abc" -1))
+  "abcdef"
+  @[])
+
+(test "sub: positions inside second match are still relative to the entire input"
+  ~(* "one\ntw" (sub "o" (* ($) (line) (column))))
+  "one\ntwo\nthree\n"
+  @[6 2 3])
+
+(test "sub: advances to the end of the first pattern's match"
+  ~(* (sub "abc" "ab") "d")
+  "abcdef"
+  @[])
+
 (end-suite)