Remove jump instructions from findPtr, rFindPtr

This results in an approx 3% deoptimisation according to benchmarks,
however allows all library functions to be marked as pure and removes the
last of the compiler warnings from the code.

Marked all library functions as pure

Author: ovangle

src/strings.sol

@@ -42,7 +42,7 @@ library strings {
         uint _ptr;
     }
 
-    function memcpy(uint dest, uint src, uint len) private {
+    function memcpy(uint dest, uint src, uint len) private pure {
         // Copy word-length chunks while possible
         for(; len >= 32; len -= 32) {
             assembly {
@@ -66,7 +66,7 @@ library strings {
      * @param self The string to make a slice from.
      * @return A newly allocated slice containing the entire string.
      */
-    function toSlice(string self) internal returns (slice) {
+    function toSlice(string self) internal pure returns (slice) {
         uint ptr;
         assembly {
             ptr := add(self, 0x20)
@@ -79,7 +79,7 @@ library strings {
      * @param self The value to find the length of.
      * @return The length of the string, from 0 to 32.
      */
-    function len(bytes32 self) internal returns (uint) {
+    function len(bytes32 self) internal pure returns (uint) {
         uint ret;
         if (self == 0)
             return 0;
@@ -112,7 +112,7 @@ library strings {
      * @return A new slice containing the value of the input argument up to the
      *         first null.
      */
-    function toSliceB32(bytes32 self) internal returns (slice ret) {
+    function toSliceB32(bytes32 self) internal pure returns (slice ret) {
         // Allocate space for `self` in memory, copy it there, and point ret at it
         assembly {
             let ptr := mload(0x40)
@@ -128,7 +128,7 @@ library strings {
      * @param self The slice to copy.
      * @return A new slice containing the same data as `self`.
      */
-    function copy(slice self) internal returns (slice) {
+    function copy(slice self) internal pure returns (slice) {
         return slice(self._len, self._ptr);
     }
 
@@ -137,7 +137,7 @@ library strings {
      * @param self The slice to copy.
      * @return A newly allocated string containing the slice's text.
      */
-    function toString(slice self) internal returns (string) {
+    function toString(slice self) internal pure returns (string) {
         string memory ret = new string(self._len);
         uint retptr;
         assembly { retptr := add(ret, 32) }
@@ -146,10 +146,6 @@ library strings {
         return ret;
     }
 
-    function len(string self) internal returns (uint l) {
-        return len(toSlice(self));
-    }
-
     /*
      * @dev Returns the length in runes of the slice. Note that this operation
      *      takes time proportional to the length of the slice; avoid using it
@@ -158,7 +154,7 @@ library strings {
      * @param self The slice to operate on.
      * @return The length of the slice in runes.
      */
-    function len(slice self) internal returns (uint l) {
+    function len(slice self) internal pure returns (uint l) {
         // Starting at ptr-31 means the LSB will be the byte we care about
         uint ptr = self._ptr - 31;
         uint end = ptr + self._len;
@@ -186,22 +182,10 @@ library strings {
      * @param self The slice to operate on.
      * @return True if the slice is empty, False otherwise.
      */
-    function empty(slice self) internal returns (bool) {
+    function empty(slice self) internal pure returns (bool) {
         return self._len == 0;
     }
 
-    function compare(string self, string other) internal returns (int) {
-        return compare(toSlice(self), toSlice(other));
-    }
-
-    function compare(string self, slice other) internal returns (int) {
-        return compare(toSlice(self), other);
-    }
-
-    function compare(slice self, string other) internal returns (int) {
-        return compare(self, toSlice(other));
-    }
-
     /*
      * @dev Returns a positive number if `other` comes lexicographically after
      *      `self`, a negative number if it comes before, or zero if the
@@ -211,7 +195,7 @@ library strings {
      * @param other The second slice to compare.
      * @return The result of the comparison.
      */
-    function compare(slice self, slice other) internal returns (int) {
+    function compare(slice self, slice other) internal pure returns (int) {
         uint shortest = self._len;
         if (other._len < self._len)
             shortest = other._len;
@@ -244,7 +228,7 @@ library strings {
      * @param self The second slice to compare.
      * @return True if the slices are equal, false otherwise.
      */
-    function equals(slice self, slice other) internal returns (bool) {
+    function equals(slice self, slice other) internal pure returns (bool) {
         return compare(self, other) == 0;
     }
 
@@ -255,7 +239,7 @@ library strings {
      * @param rune The slice that will contain the first rune.
      * @return `rune`.
      */
-    function nextRune(slice self, slice rune) internal returns (slice) {
+    function nextRune(slice self, slice rune) internal pure returns (slice) {
         rune._ptr = self._ptr;
 
         if (self._len == 0) {
@@ -297,7 +281,7 @@ library strings {
      * @param self The slice to operate on.
      * @return A slice containing only the first rune from `self`.
      */
-    function nextRune(slice self) internal returns (slice ret) {
+    function nextRune(slice self) internal pure returns (slice ret) {
         nextRune(self, ret);
     }
 
@@ -306,7 +290,7 @@ library strings {
      * @param self The slice to operate on.
      * @return The number of the first codepoint in the slice.
      */
-    function ord(slice self) internal returns (uint ret) {
+    function ord(slice self) internal pure returns (uint ret) {
         if (self._len == 0) {
             return 0;
         }
@@ -355,7 +339,7 @@ library strings {
      * @param self The slice to hash.
      * @return The hash of the slice.
      */
-    function keccak(slice self) internal returns (bytes32 ret) {
+    function keccak(slice self) internal pure returns (bytes32 ret) {
         assembly {
             ret := keccak256(mload(add(self, 32)), mload(self))
         }
@@ -367,7 +351,7 @@ library strings {
      * @param needle The slice to search for.
      * @return True if the slice starts with the provided text, false otherwise.
      */
-    function startsWith(slice self, slice needle) internal returns (bool) {
+    function startsWith(slice self, slice needle) internal pure returns (bool) {
         if (self._len < needle._len) {
             return false;
         }
@@ -393,7 +377,7 @@ library strings {
      * @param needle The slice to search for.
      * @return `self`
      */
-    function beyond(slice self, slice needle) internal returns (slice) {
+    function beyond(slice self, slice needle) internal pure returns (slice) {
         if (self._len < needle._len) {
             return self;
         }
@@ -422,7 +406,7 @@ library strings {
      * @param needle The slice to search for.
      * @return True if the slice starts with the provided text, false otherwise.
      */
-    function endsWith(slice self, slice needle) internal returns (bool) {
+    function endsWith(slice self, slice needle) internal pure returns (bool) {
         if (self._len < needle._len) {
             return false;
         }
@@ -450,7 +434,7 @@ library strings {
      * @param needle The slice to search for.
      * @return `self`
      */
-    function until(slice self, slice needle) internal returns (slice) {
+    function until(slice self, slice needle) internal pure returns (slice) {
         if (self._len < needle._len) {
             return self;
         }
@@ -472,33 +456,37 @@ library strings {
         return self;
     }
 
+    event log_bytemask(bytes32 mask);
+
     // Returns the memory address of the first byte of the first occurrence of
     // `needle` in `self`, or the first byte after `self` if not found.
-    function findPtr(uint selflen, uint selfptr, uint needlelen, uint needleptr) private returns (uint) {
-        uint ptr;
+    function findPtr(uint selflen, uint selfptr, uint needlelen, uint needleptr) private pure returns (uint) {
+        uint ptr = selfptr;
         uint idx;
 
         if (needlelen <= selflen) {
             if (needlelen <= 32) {
-                // Optimized assembly for 68 gas per byte on short strings
-                assembly {
-                    let mask := not(sub(exp(2, mul(8, sub(32, needlelen))), 1))
-                    let needledata := and(mload(needleptr), mask)
-                    let end := add(selfptr, sub(selflen, needlelen))
-                    ptr := selfptr
-                    loop:
-                    jumpi(exit, eq(and(mload(ptr), mask), needledata))
-                    ptr := add(ptr, 1)
-                    jumpi(loop, lt(sub(ptr, 1), end))
-                    ptr := add(selfptr, selflen)
-                    exit:
+                bytes32 mask = bytes32(~(2 ** (8 * (32 - needlelen)) - 1));
+
+                bytes32 needledata;
+                assembly { needledata := and(mload(needleptr), mask) }
+
+                uint end = selfptr + selflen - needlelen;
+                bytes32 ptrdata;
+                assembly { ptrdata := and(mload(ptr), mask) }
+
+                while (ptrdata != needledata) {
+                    if (ptr >= end) 
+                        return selfptr + selflen;
+                    ptr++;
+                    assembly { ptrdata := and(mload(ptr), mask) }
                 }
                 return ptr;
             } else {
                 // For long needles, use hashing
                 bytes32 hash;
                 assembly { hash := sha3(needleptr, needlelen) }
-                ptr = selfptr;
+
                 for (idx = 0; idx <= selflen - needlelen; idx++) {
                     bytes32 testHash;
                     assembly { testHash := sha3(ptr, needlelen) }
@@ -513,27 +501,27 @@ library strings {
 
     // Returns the memory address of the first byte after the last occurrence of
     // `needle` in `self`, or the address of `self` if not found.
-    function rfindPtr(uint selflen, uint selfptr, uint needlelen, uint needleptr) private returns (uint) {
+    function rfindPtr(uint selflen, uint selfptr, uint needlelen, uint needleptr) private pure returns (uint) {
         uint ptr;
 
         if (needlelen <= selflen) {
             if (needlelen <= 32) {
-                // Optimized assembly for 69 gas per byte on short strings
-                assembly {
-                    let mask := not(sub(exp(2, mul(8, sub(32, needlelen))), 1))
-                    let needledata := and(mload(needleptr), mask)
-                    ptr := add(selfptr, sub(selflen, needlelen))
-                    loop:
-                    jumpi(ret, eq(and(mload(ptr), mask), needledata))
-                    ptr := sub(ptr, 1)
-                    jumpi(loop, gt(add(ptr, 1), selfptr))
-                    ptr := selfptr
-                    jump(exit)
-                    ret:
-                    ptr := add(ptr, needlelen)
-                    exit:
+                bytes32 mask = bytes32(~(2 ** (8 * (32 - needlelen)) - 1));
+
+                bytes32 needledata;
+                assembly { needledata := and(mload(needleptr), mask) }
+
+                ptr = selfptr + selflen - needlelen;
+                bytes32 ptrdata;
+                assembly { ptrdata := and(mload(ptr), mask) }
+
+                while (ptrdata != needledata) {
+                    if (ptr <= selfptr) 
+                        return selfptr;
+                    ptr--;
+                    assembly { ptrdata := and(mload(ptr), mask) }
                 }
-                return ptr;
+                return ptr + needlelen;
             } else {
                 // For long needles, use hashing
                 bytes32 hash;
@@ -559,7 +547,7 @@ library strings {
      * @param needle The text to search for.
      * @return `self`.
      */
-    function find(slice self, slice needle) internal returns (slice) {
+    function find(slice self, slice needle) internal pure returns (slice) {
         uint ptr = findPtr(self._len, self._ptr, needle._len, needle._ptr);
         self._len -= ptr - self._ptr;
         self._ptr = ptr;
@@ -574,7 +562,7 @@ library strings {
      * @param needle The text to search for.
      * @return `self`.
      */
-    function rfind(slice self, slice needle) internal returns (slice) {
+    function rfind(slice self, slice needle) internal pure returns (slice) {
         uint ptr = rfindPtr(self._len, self._ptr, needle._len, needle._ptr);
         self._len = ptr - self._ptr;
         return self;
@@ -590,7 +578,7 @@ library strings {
      * @param token An output parameter to which the first token is written.
      * @return `token`.
      */
-    function split(slice self, slice needle, slice token) internal returns (slice) {
+    function split(slice self, slice needle, slice token) internal pure returns (slice) {
         uint ptr = findPtr(self._len, self._ptr, needle._len, needle._ptr);
         token._ptr = self._ptr;
         token._len = ptr - self._ptr;
@@ -613,7 +601,7 @@ library strings {
      * @param needle The text to search for in `self`.
      * @return The part of `self` up to the first occurrence of `delim`.
      */
-    function split(slice self, slice needle) internal returns (slice token) {
+    function split(slice self, slice needle) internal pure returns (slice token) {
         split(self, needle, token);
     }
 
@@ -627,7 +615,7 @@ library strings {
      * @param token An output parameter to which the first token is written.
      * @return `token`.
      */
-    function rsplit(slice self, slice needle, slice token) internal returns (slice) {
+    function rsplit(slice self, slice needle, slice token) internal pure returns (slice) {
         uint ptr = rfindPtr(self._len, self._ptr, needle._len, needle._ptr);
         token._ptr = ptr;
         token._len = self._len - (ptr - self._ptr);
@@ -649,7 +637,7 @@ library strings {
      * @param needle The text to search for in `self`.
      * @return The part of `self` after the last occurrence of `delim`.
      */
-    function rsplit(slice self, slice needle) internal returns (slice token) {
+    function rsplit(slice self, slice needle) internal pure returns (slice token) {
         rsplit(self, needle, token);
     }
 
@@ -659,7 +647,7 @@ library strings {
      * @param needle The text to search for in `self`.
      * @return The number of occurrences of `needle` found in `self`.
      */
-    function count(slice self, slice needle) internal returns (uint cnt) {
+    function count(slice self, slice needle) internal pure returns (uint cnt) {
         uint ptr = findPtr(self._len, self._ptr, needle._len, needle._ptr) + needle._len;
         while (ptr <= self._ptr + self._len) {
             cnt++;
@@ -673,7 +661,7 @@ library strings {
      * @param needle The text to search for in `self`.
      * @return True if `needle` is found in `self`, false otherwise.
      */
-    function contains(slice self, slice needle) internal returns (bool) {
+    function contains(slice self, slice needle) internal pure returns (bool) {
         return rfindPtr(self._len, self._ptr, needle._len, needle._ptr) != self._ptr;
     }
 
@@ -684,7 +672,7 @@ library strings {
      * @param other The second slice to concatenate.
      * @return The concatenation of the two strings.
      */
-    function concat(slice self, slice other) internal returns (string) {
+    function concat(slice self, slice other) internal pure returns (string) {
         string memory ret = new string(self._len + other._len);
         uint retptr;
         assembly { retptr := add(ret, 32) }
@@ -701,7 +689,7 @@ library strings {
      * @return A newly allocated string containing all the slices in `parts`,
      *         joined with `self`.
      */
-    function join(slice self, slice[] parts) internal returns (string) {
+    function join(slice self, slice[] parts) internal pure returns (string) {
         if (parts.length == 0)
             return "";
 

src/strings.t.sol

@@ -6,7 +6,6 @@ import './strings.sol';
 contract StringsTest is DSTest {
 	using strings for *;
 
-    event log_named_string(bytes32 key, string value);
 
     function abs(int x) private pure returns (int) {
         if(x < 0)