commit 26bc1afb8dedd15261634fe2104e79a7ccdcd21b
author Paul Wankadia <junyer@google.com> Tue May 19 08:31:19 2020 -0700
committer Paul Wankadia <junyer@google.com> Tue May 19 15:32:52 2020 +0000
tree c0a7dd5ae04194eec502697c84d2db0d843c2f67
parent b9e7f273fd1aae2a83324d10a253165fcdd33ece

Tidy up the code around the memchr(3) calls.

Change-Id: Id06cae00beaebeaa74f59d68a9abc5e39e37ced6
Reviewed-on: https://code-review.googlesource.com/c/re2/+/56650
Reviewed-by: Paul Wankadia <junyer@google.com>

re2/bitstate.cc

diff --git a/re2/bitstate.cc b/re2/bitstate.cc
index 865b58e..0f7ab47 100644
--- a/re2/bitstate.cc
+++ b/re2/bitstate.cc
@@ -332,14 +332,14 @@
   // This looks like it's quadratic in the size of the text,
   // but we are not clearing visited_ between calls to TrySearch,
   // so no work is duplicated and it ends up still being linear.
-  for (const char* p = text.data(); p <= text.data() + text.size(); p++) {
+  const char* etext = text.data() + text.size();
+  for (const char* p = text.data(); p <= etext; p++) {
     // Try to use memchr to find the first byte quickly.
-    int fb = prog_->first_byte();
-    if (fb >= 0 && p < text.data() + text.size() && (p[0] & 0xFF) != fb) {
-      p = reinterpret_cast<const char*>(
-          memchr(p, fb, text.data() + text.size() - p));
+    int first_byte = prog_->first_byte();
+    if (first_byte >= 0 && p < etext && (p[0] & 0xFF) != first_byte) {
+      p = reinterpret_cast<const char*>(memchr(p, first_byte, etext - p));
       if (p == NULL)
-        p = text.data() + text.size();
+        p = etext;
     }
 
     cap_[0] = p;

re2/dfa.cc

diff --git a/re2/dfa.cc b/re2/dfa.cc
index 9e2276e..6581aec 100644
--- a/re2/dfa.cc
+++ b/re2/dfa.cc
@@ -1337,7 +1337,6 @@
     swap(p, ep);
   }
 
-  int first_byte = prog_->first_byte();
   const uint8_t* bytemap = prog_->bytemap();
   const uint8_t* lastmatch = NULL;   // most recent matching position in text
   bool matched = false;
@@ -1374,7 +1373,9 @@
       // so use optimized assembly in memchr to skip ahead.
       // If first_byte isn't found, we can skip to the end
       // of the string.
-      if ((p = BytePtr(memchr(p, first_byte, ep - p))) == NULL) {
+      int first_byte = prog_->first_byte();
+      p = BytePtr(memchr(p, first_byte, ep - p));
+      if (p == NULL) {
         p = ep;
         break;
       }

re2/nfa.cc

diff --git a/re2/nfa.cc b/re2/nfa.cc
index 6ef918f..8e0ace9 100644
--- a/re2/nfa.cc
+++ b/re2/nfa.cc
@@ -574,13 +574,12 @@
       // If there's a required first byte for an unanchored search
       // and we're not in the middle of any possible matches,
       // use memchr to search for the byte quickly.
-      int fb = prog_->first_byte();
+      int first_byte = prog_->first_byte();
       if (!anchored && runq->size() == 0 &&
-          fb >= 0 && p < etext_ && (p[0] & 0xFF) != fb) {
-        p = reinterpret_cast<const char*>(memchr(p, fb, etext_ - p));
-        if (p == NULL) {
+          first_byte >= 0 && p < etext_ && (p[0] & 0xFF) != first_byte) {
+        p = reinterpret_cast<const char*>(memchr(p, first_byte, etext_ - p));
+        if (p == NULL)
           p = etext_;
-        }
       }
 
       Thread* t = AllocThread();