From b0f931f18d4ce98ebbc712d0caa8368b68137028 Mon Sep 17 00:00:00 2001
From: Evan Martin <martine@danga.com>
Date: Fri, 17 Dec 2010 15:00:47 -0800
Subject: use hash_map for paths; much faster builds

---
 src/eval_env.h |  3 +++
 src/hash_map.h | 12 ++++++++++++
 src/ninja.h    |  4 ++--
 3 files changed, 17 insertions(+), 2 deletions(-)
 create mode 100644 src/hash_map.h

diff --git a/src/eval_env.h b/src/eval_env.h
index a630e7a..cc61a38 100644
--- a/src/eval_env.h
+++ b/src/eval_env.h
@@ -1,6 +1,9 @@
 #ifndef NINJA_EVAL_ENV_H_
 #define NINJA_EVAL_ENV_H_
 
+#include <map>
+using namespace std;
+
 // A scope for variable lookups.
 struct Env {
   virtual string LookupVariable(const string& var) = 0;
diff --git a/src/hash_map.h b/src/hash_map.h
new file mode 100644
index 0000000..820f773
--- /dev/null
+++ b/src/hash_map.h
@@ -0,0 +1,12 @@
+#include <ext/hash_map>
+
+using __gnu_cxx::hash_map;
+
+namespace __gnu_cxx {
+template<>
+struct hash<std::string> {
+  size_t operator()(const std::string& s) const {
+    return hash<const char*>()(s.c_str());
+  }
+};
+}
diff --git a/src/ninja.h b/src/ninja.h
index ecfb2d0..fb0b7fb 100644
--- a/src/ninja.h
+++ b/src/ninja.h
@@ -2,7 +2,6 @@
 #define NINJA_NINJA_H_
 
 #include <algorithm>
-#include <map>
 #include <queue>
 #include <set>
 #include <string>
@@ -13,6 +12,7 @@
 using namespace std;
 
 #include "eval_env.h"
+#include "hash_map.h"
 
 int ReadFile(const string& path, string* contents, string* err);
 
@@ -135,7 +135,7 @@ struct Edge {
 };
 
 struct StatCache {
-  typedef map<string, FileStat*> Paths;
+  typedef hash_map<string, FileStat*> Paths;
   Paths paths_;
   FileStat* GetFile(const string& path);
   void Dump();
-- 
cgit v0.12


From 3261f368ba73ece86d0fb190bc6205ca682e1ae3 Mon Sep 17 00:00:00 2001
From: Evan Martin <martine@danga.com>
Date: Fri, 17 Dec 2010 15:21:13 -0800
Subject: simplify token representation, speeding up parse

---
 src/parsers.cc | 17 +++++++++--------
 src/parsers.h  |  4 ++--
 2 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/src/parsers.cc b/src/parsers.cc
index 4d6c058..4041bf6 100644
--- a/src/parsers.cc
+++ b/src/parsers.cc
@@ -9,8 +9,8 @@
 
 string Token::AsString() const {
   switch (type_) {
-  case IDENT:    return "'" + extra_ + "'";
-  case UNKNOWN:  return "unknown '" + extra_ + "'";
+  case IDENT:    return "'" + string(pos_, end_ - pos_) + "'";
+  case UNKNOWN:  return "unknown '" + string(pos_, end_ - pos_) + "'";
   case RULE:     return "'rule'";
   case BUILD:    return "'build'";
   case SUBNINJA: return "'subninja'";
@@ -99,7 +99,7 @@ bool Tokenizer::ReadIdent(string* out) {
   PeekToken();
   if (token_.type_ != Token::IDENT)
     return false;
-  out->assign(token_.extra_);
+  out->assign(token_.pos_, token_.end_ - token_.pos_);
   ConsumeToken();
   return true;
 }
@@ -159,14 +159,15 @@ Token::Type Tokenizer::PeekToken() {
 
   if (IsIdentChar(*cur_)) {
     while (cur_ < end_ && IsIdentChar(*cur_)) {
-      token_.extra_.push_back(*cur_);
       ++cur_;
     }
-    if (token_.extra_ == "rule")
+    token_.end_ = cur_;
+    int len = token_.end_ - token_.pos_;
+    if (len == 4 && memcmp(token_.pos_, "rule", 4) == 0)
       token_.type_ = Token::RULE;
-    else if (token_.extra_ == "build")
+    else if (len == 5 && memcmp(token_.pos_, "build", 5) == 0)
       token_.type_ = Token::BUILD;
-    else if (token_.extra_ == "subninja")
+    else if (len == 8 && memcmp(token_.pos_, "subninja", 8) == 0)
       token_.type_ = Token::SUBNINJA;
     else
       token_.type_ = Token::IDENT;
@@ -191,7 +192,7 @@ Token::Type Tokenizer::PeekToken() {
 
   if (token_.type_ == Token::NONE) {
     token_.type_ = Token::UNKNOWN;
-    token_.extra_ = *cur_;
+    token_.end_ = cur_ + 1;
   }
 
   return token_.type_;
diff --git a/src/parsers.h b/src/parsers.h
index fc20436..a7f23e5 100644
--- a/src/parsers.h
+++ b/src/parsers.h
@@ -26,12 +26,12 @@ struct Token {
   };
   explicit Token(Type type) : type_(type) {}
 
-  void Clear() { type_ = NONE; extra_.clear(); }
+  void Clear() { type_ = NONE; }
   string AsString() const;
 
   Type type_;
   const char* pos_;
-  string extra_;
+  const char* end_;
 };
 
 struct Tokenizer {
-- 
cgit v0.12