reimplement isPrintableValidUTF8 using phobos functions

This commit is contained in:
Cameron Ross 2019-02-04 20:52:04 -04:00
parent b63ea1aaae
commit ff38f20b09
No known key found for this signature in database
GPG key ID: 777897D98DC91C54

View file

@ -803,109 +803,10 @@ auto toUTF8(ubyte[] input, const UTFEncoding encoding) @safe pure nothrow
/// Determine if all characters (code points, not bytes) in a string are printable. /// Determine if all characters (code points, not bytes) in a string are printable.
bool isPrintableValidUTF8(const char[] chars) @safe pure bool isPrintableValidUTF8(const char[] chars) @safe pure
{ {
// This is oversized (only 128 entries are necessary) simply because having 256 import std.uni : isControl, isWhite;
// entries improves performance... for some reason (alignment?) foreach (dchar chr; chars)
bool[256] printable = [false, false, false, false, false, false, false, false,
false, true, true, false, false, true, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
true, true, true, true, true, true, true, true,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false,
false, false, false, false, false, false, false, false];
for(size_t index; index < chars.length;)
{ {
// Fast path for ASCII. if (!chr.isValidDchar || (chr.isControl && !chr.isWhite))
// Both this while() block and the if() block below it are optimized, unrolled
// versions of the for() block below them; the while()/if() block could be
// removed without affecting logic, but both help increase performance.
size_t asciiCount = countASCII(chars[index .. $]);
// 8 ASCII iterations unrolled, looping while there are at most 8 ASCII chars.
while(asciiCount > 8)
{
const dchar b0 = chars[index];
const dchar b1 = chars[index + 1];
const dchar b2 = chars[index + 2];
const dchar b3 = chars[index + 3];
const dchar b4 = chars[index + 4];
const dchar b5 = chars[index + 5];
const dchar b6 = chars[index + 6];
const dchar b7 = chars[index + 7];
index += 8;
asciiCount -= 8;
const all = printable[b0] & printable[b1] & printable[b2] & printable[b3] &
printable[b4] & printable[b5] & printable[b6] & printable[b1];
if(!all)
{
return false;
}
}
// 4 ASCII iterations unrolled
if(asciiCount > 4)
{
const char b0 = chars[index];
const char b1 = chars[index + 1];
const char b2 = chars[index + 2];
const char b3 = chars[index + 3];
index += 4;
asciiCount -= 4;
if(!printable[b0]) { return false; }
if(!printable[b1]) { return false; }
if(!printable[b2]) { return false; }
if(!printable[b3]) { return false; }
}
// Any remaining ASCII chars. This is really the only code needed to handle
// ASCII, the above if() and while() blocks are just an optimization.
for(; asciiCount > 0; --asciiCount)
{
const char b = chars[index];
++index;
if(b >= 0x20) { continue; }
if(printable[b]) { continue; }
return false;
}
if(index == chars.length) { break; }
// Not ASCII, need to decode.
const dchar c = decode(chars, index);
// We now c is not ASCII, so only check for printable non-ASCII chars.
if(!(c == 0x85 || (c >= 0xA0 && c <= '\uD7FF') ||
(c >= '\uE000' && c <= '\uFFFD') ||
(c >= '\U00010000' && c <= '\U0010FFFF')))
{ {
return false; return false;
} }