Improve scan codegen

src/parser3.h

@@ -16,6 +16,316 @@
 
 namespace parser3 {
 
+struct NumDfa {
+  constexpr static uint64_t num_dfa_table[256] = {
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x36000ull,
+      0x0ull,
+      0x36600ull,
+      0x12480000000000ull,
+      0x0ull,
+      0x780aa47b091ec00ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x780aa47aa91ea80ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0xc30c000000000ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0xc30c000000000ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+      0x0ull,
+  };
+  // Restore this dfa to its start state
+  void reset() { state = 6; }
+  // Return true if this dfa is in an accept state. You probably want to call
+  // scan until the match ends first.
+  bool accept() const {
+    return (state & 63) == 30 || (state & 63) == 36 || (state & 63) == 48 ||
+           (state & 63) == 42;
+  }
+  // return value either points to the first byte which does not match, or
+  // bufEnd. Leaves the dfa in the last state of the match.
+#ifdef __x86_64__
+  __attribute__((target_clones("default", "bmi2")))
+#endif
+  const char *
+  scan(const char *buf, const char *bufEnd) {
+    auto state_ = state;
+    for (;;) {
+      constexpr int kStride = 16;
+      if (bufEnd - buf < kStride) [[unlikely]] {
+        while (buf != bufEnd) {
+          uint64_t row = num_dfa_table[uint8_t(*buf)];
+          auto prev = state_;
+          state_ = (row >> (state_ & 63)) & 63;
+          if (state_ == 0) {
+            state_ = prev;
+            break;
+          }
+          ++buf;
+        }
+        state = state_;
+        return buf;
+      }
+      uint8_t prev[kStride + 1];
+      prev[0] = state_;
+      for (int i = 0; i < kStride; ++i) {
+        uint64_t row = num_dfa_table[uint8_t(*buf)];
+        prev[i + 1] = row >> (prev[i] & 63);
+        if ((prev[i + 1] & 63) == 0) {
+          state = prev[i];
+          return buf;
+        }
+        ++buf;
+      }
+      state_ = prev[kStride];
+    }
+  }
+
+private:
+  uint64_t state = 6;
+};
+
 typedef PRESERVE_NONE WeaselJsonStatus (*Continuation)(struct Parser3 *,
                                                        char *buf, char *bufEnd);
 
@@ -30,13 +340,8 @@ enum Symbol : uint8_t {
   N_STRING,
   N_STRING2,
   N_STRING_FOLLOWING_ESCAPE,
-  N_INTEGER2,
-  N_DIGITS,
-  N_DIGITS2,
-  N_FRACTION,
-  N_EXPONENT,
-  N_SIGN,
   N_WHITESPACE,
+  N_NUMBER,
   N_TRUE,
   N_FALSE,
   N_NULL,
@@ -53,10 +358,7 @@ enum Symbol : uint8_t {
   T_HEX,
   T_HEX2,
   T_HEX3,
-  T_DIGIT,
-  T_ONENINE,
   T_EOF,
-  T_END_NUMBER,
   T_BACKSLASH,
   N_SYMBOL_COUNT, // Must be last
 };
@@ -131,6 +433,7 @@ struct Parser3 {
   uint32_t minCodepoint;
   int const stackSize;
   bool complete;
+  NumDfa numDfa;
 };
 
 inline PRESERVE_NONE WeaselJsonStatus n_whitespace(Parser3 *self, char *buf,
@@ -149,6 +452,21 @@ inline PRESERVE_NONE WeaselJsonStatus n_whitespace(Parser3 *self, char *buf,
   MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
 }
 
+inline PRESERVE_NONE WeaselJsonStatus n_number(Parser3 *self, char *buf,
+                                               char *bufEnd) {
+  buf = (char *)self->numDfa.scan(buf, bufEnd);
+  if (buf == bufEnd && !self->complete) {
+    self->flushNumber(false, buf);
+    return WeaselJson_AGAIN;
+  }
+  if (!self->numDfa.accept()) [[unlikely]] {
+    return WeaselJson_REJECT;
+  }
+  self->flushNumber(true, buf);
+  self->pop();
+  MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
+}
+
 inline PRESERVE_NONE WeaselJsonStatus n_value(Parser3 *self, char *buf,
                                               char *bufEnd) {
   assert(bufEnd - buf != 0);
@@ -184,13 +502,6 @@ inline PRESERVE_NONE WeaselJsonStatus n_value(Parser3 *self, char *buf,
     }
     break;
   case '0':
-    self->pop();
-    if (auto s = self->push({N_FRACTION, N_EXPONENT})) {
-      return s;
-    }
-    self->dataBegin = buf;
-    ++buf;
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
   case '1':
   case '2':
   case '3':
@@ -200,20 +511,22 @@ inline PRESERVE_NONE WeaselJsonStatus n_value(Parser3 *self, char *buf,
   case '7':
   case '8':
   case '9':
-    self->pop();
-    self->dataBegin = buf;
-    ++buf;
-    if (auto s = self->push({N_DIGITS2, N_FRACTION, N_EXPONENT})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
   case '-':
-    self->pop();
     self->dataBegin = buf;
-    ++buf;
+    self->pop();
-    if (auto s = self->push({N_INTEGER2, N_FRACTION, N_EXPONENT})) {
+    self->numDfa.reset();
-      return s;
+    buf = (char *)self->numDfa.scan(buf, bufEnd);
+    if (buf == bufEnd) {
+      self->flushNumber(false, buf);
+      if (auto s = self->push({N_NUMBER})) {
+        return s;
+      }
+      return WeaselJson_AGAIN;
     }
+    if (!self->numDfa.accept()) [[unlikely]] {
+      return WeaselJson_REJECT;
+    }
+    self->flushNumber(true, buf);
     MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
   case 't':
     ++buf;
@@ -733,171 +1046,6 @@ inline PRESERVE_NONE WeaselJsonStatus t_hex3(Parser3 *self, char *buf,
   MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
 }
 
-inline PRESERVE_NONE WeaselJsonStatus n_integer(Parser3 *self, char *buf,
-                                                char *bufEnd) {
-  self->dataBegin = buf;
-  switch (*buf) {
-  case '0':
-    ++buf;
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  case '1':
-  case '2':
-  case '3':
-  case '4':
-  case '5':
-  case '6':
-  case '7':
-  case '8':
-  case '9':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_DIGITS2})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  case '-':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_INTEGER2})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    [[unlikely]] return WeaselJson_REJECT;
-  }
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_integer2(Parser3 *self, char *buf,
-                                                 char *bufEnd) {
-  switch (*buf) {
-  case '0':
-    ++buf;
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  case '1':
-  case '2':
-  case '3':
-  case '4':
-  case '5':
-  case '6':
-  case '7':
-  case '8':
-  case '9':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_DIGITS2})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    [[unlikely]] return WeaselJson_REJECT;
-  }
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_digits(Parser3 *self, char *buf,
-                                               char *bufEnd) {
-  switch (*buf) {
-  case '0':
-  case '1':
-  case '2':
-  case '3':
-  case '4':
-  case '5':
-  case '6':
-  case '7':
-  case '8':
-  case '9':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_DIGITS2})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    [[unlikely]] return WeaselJson_REJECT;
-  }
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_digits2(Parser3 *self, char *buf,
-                                                char *bufEnd) {
-  if (bufEnd - buf == 0) {
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-
-  // Advance buf to the first non-decimal character
-  while (buf != bufEnd && '0' <= *buf && *buf <= '9') {
-    ++buf;
-  }
-  if (buf == bufEnd) {
-    self->flushNumber(false, buf);
-    return WeaselJson_AGAIN;
-  }
-  self->pop();
-  MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_fraction(Parser3 *self, char *buf,
-                                                 char *bufEnd) {
-  if (bufEnd - buf == 0) {
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-  switch (*buf) {
-  case '.':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_DIGITS})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_exponent(Parser3 *self, char *buf,
-                                                 char *bufEnd) {
-  if (bufEnd - buf == 0) {
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-  switch (*buf) {
-  case 'e':
-  case 'E':
-    ++buf;
-    self->pop();
-    if (auto s = self->push({N_SIGN, N_DIGITS, T_END_NUMBER})) {
-      return s;
-    }
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    self->pop();
-    self->flushNumber(true, buf);
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-}
-
-inline PRESERVE_NONE WeaselJsonStatus n_sign(Parser3 *self, char *buf,
-                                             char *bufEnd) {
-  if (bufEnd - buf == 0) {
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-  switch (*buf) {
-  case '+':
-  case '-':
-    ++buf;
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  default:
-    self->pop();
-    MUSTTAIL return Parser3::keepGoing(self, buf, bufEnd);
-  }
-}
-
 inline PRESERVE_NONE WeaselJsonStatus n_true(Parser3 *self, char *buf,
                                              char *bufEnd) {
   if (*buf == 'e') {
@@ -987,13 +1135,8 @@ constexpr inline struct ContinuationTable {
     continuations[N_STRING] = n_string;
     continuations[N_STRING2] = n_string2;
     continuations[N_STRING_FOLLOWING_ESCAPE] = n_string_following_escape;
-    continuations[N_INTEGER2] = n_integer2;
-    continuations[N_DIGITS] = n_digits;
-    continuations[N_DIGITS2] = n_digits2;
-    continuations[N_FRACTION] = n_fraction;
-    continuations[N_EXPONENT] = n_exponent;
-    continuations[N_SIGN] = n_sign;
     continuations[N_WHITESPACE] = n_whitespace;
+    continuations[N_NUMBER] = n_number;
     continuations[N_TRUE] = n_true;
     continuations[N_FALSE] = n_false;
     continuations[N_NULL] = n_null;
@@ -1010,10 +1153,7 @@ constexpr inline struct ContinuationTable {
     continuations[T_HEX] = t_hex;
     continuations[T_HEX2] = t_hex2;
     continuations[T_HEX3] = t_hex3;
-    continuations[T_DIGIT] = t_digit;
-    continuations[T_ONENINE] = t_onenine;
     continuations[T_EOF] = t_eof;
-    continuations[T_END_NUMBER] = t_end_number;
     continuations[T_BACKSLASH] = singleChar<'\\'>;
 
     symbolNames[N_VALUE] = "n_value";
@@ -1024,13 +1164,8 @@ constexpr inline struct ContinuationTable {
     symbolNames[N_STRING] = "n_string";
     symbolNames[N_STRING2] = "n_string2";
     symbolNames[N_STRING_FOLLOWING_ESCAPE] = "n_string_following_escape";
-    symbolNames[N_INTEGER2] = "n_integer2";
-    symbolNames[N_DIGITS] = "n_digits";
-    symbolNames[N_DIGITS2] = "n_digits2";
-    symbolNames[N_FRACTION] = "n_fraction";
-    symbolNames[N_EXPONENT] = "n_exponent";
-    symbolNames[N_SIGN] = "n_sign";
     symbolNames[N_WHITESPACE] = "n_whitespace";
+    symbolNames[N_NUMBER] = "n_number";
     symbolNames[N_TRUE] = "n_true";
     symbolNames[N_FALSE] = "n_false";
     symbolNames[N_NULL] = "n_null";
@@ -1045,20 +1180,13 @@ constexpr inline struct ContinuationTable {
     symbolNames[T_HEX] = "t_hex";
     symbolNames[T_HEX2] = "t_hex2";
     symbolNames[T_HEX3] = "t_hex3";
-    symbolNames[T_DIGIT] = "t_digit";
-    symbolNames[T_ONENINE] = "t_onenine";
     symbolNames[T_EOF] = "t_eof";
     symbolNames[T_BACKSLASH] = "singleChar<'\\'>";
-    symbolNames[T_END_NUMBER] = "t_end_number";
 
     // All others can assume that there's at least one byte when they're called
-    acceptsEmptyString[N_DIGITS2] = true;
+    acceptsEmptyString[N_NUMBER] = true;
-    acceptsEmptyString[N_FRACTION] = true;
-    acceptsEmptyString[N_EXPONENT] = true;
-    acceptsEmptyString[N_SIGN] = true;
     acceptsEmptyString[N_WHITESPACE] = true;
     acceptsEmptyString[T_EOF] = true;
-    acceptsEmptyString[T_END_NUMBER] = true;
   }
   Continuation continuations[N_SYMBOL_COUNT]{};
   const char *symbolNames[N_SYMBOL_COUNT]{};
@@ -1071,17 +1199,6 @@ inline PRESERVE_NONE WeaselJsonStatus Parser3::keepGoing(Parser3 *self,
   if (bufEnd - buf == 0) {
     if (!self->complete) {
       switch (self->top()) {
-      case N_INTEGER2:
-      case N_DIGITS:
-      case N_DIGITS2:
-      case N_FRACTION:
-      case N_EXPONENT:
-      case N_SIGN:
-      case T_DIGIT:
-      case T_ONENINE:
-      case T_END_NUMBER:
-        self->flushNumber(false, buf);
-        break;
       case N_STRING2:
       case N_STRING_FOLLOWING_ESCAPE:
       case T_UTF8_CONTINUATION_BYTE:
@@ -1101,6 +1218,7 @@ inline PRESERVE_NONE WeaselJsonStatus Parser3::keepGoing(Parser3 *self,
       case N_ARRAY2:
       case N_ARRAY3:
       case N_WHITESPACE:
+      case N_NUMBER:
       case N_TRUE:
       case N_FALSE:
       case N_NULL:

src/test.cpp

@@ -12,6 +12,7 @@
 #include <simdjson.h>
 
 #include "callbacks.h"
+#include "parser3.h"
 #include "weaseljson.h"
 
 // This is the JSON grammar in McKeeman Form.
@@ -306,3 +307,21 @@ TEST_CASE("bench5") {
     bench.doNotOptimizeAway(doc);
   });
 }
+
+TEST_CASE("num dfa") {
+  parser3::NumDfa dfa;
+  std::string match = "-1231279127389127389127398127389712893791287389217327482"
+                      "374.0e69010101010101010101010101010101";
+  auto *buf = dfa.scan(match.data(), match.data() + match.size());
+  CHECK(buf == match.data() + match.size());
+  CHECK(dfa.accept());
+
+  ankerl::nanobench::Bench bench;
+  bench.batch(match.size());
+  bench.unit("byte");
+  bench.run("number dfa", [&]() {
+    dfa.reset();
+    bench.doNotOptimizeAway(
+        dfa.scan(match.data(), match.data() + match.size()));
+  });
+}