1 module nxt.bylinefast; 2 3 import std.stdio : File; 4 import std.typecons : Flag; 5 6 alias KeepTerminator = Flag!"keepTerminator"; 7 8 /** 9 Reads by line in an efficient way (10 times faster than File.byLine from 10 std.stdio). This is accomplished by reading entire buffers (fgetc() is not 11 used), and allocating as little as possible. 12 13 The char \n is considered as default separator, removing the previous \r if 14 it exists. 15 16 The \n is never returned. The \r is not returned if it was 17 part of a \r\n (but it is returned if it was by itself). 18 19 The returned string is always a substring of a temporary buffer, that must 20 not be stored. If necessary, you must use str[] or .dup or .idup to copy to 21 another string. DIP-25 return qualifier is used in front() to add extra 22 checks in @safe callers of front(). 23 24 Example: 25 26 File f = File("file.txt"); 27 foreach (string line; ByLineFast(f)) { 28 ...process line... 29 //Make a copy: 30 string copy = line[]; 31 } 32 33 The file isn't closed when done iterating, unless it was the only reference to 34 the file (same as std.stdio.byLine). (example: ByLineFast(File("file.txt"))). 35 */ 36 struct ByLineFast(Char, Terminator) 37 { 38 import core.stdc.string : memmove; 39 import std.stdio : fgetc, ungetc; 40 41 File file; 42 char[] line; 43 bool first_call = true; 44 char[] buffer; 45 char[] strBuffer; 46 const string separator; 47 KeepTerminator keepTerminator; 48 49 this(File f, 50 KeepTerminator kt = KeepTerminator.no, 51 string separator = "\n", 52 uint bufferSize = 4096) @safe 53 { 54 assert(bufferSize > 0); 55 file = f; 56 this.separator = separator; 57 this.keepTerminator = kt; 58 buffer.length = bufferSize; 59 } 60 61 @property bool empty() const @trusted scope 62 { 63 // Its important to check "line !is null" instead of 64 // "line.length != 0", otherwise, no empty lines can 65 // be returned, the iteration would be closed. 66 if (line !is null) 67 { 68 return false; 69 } 70 if (!file.isOpen) 71 { 72 // Clean the buffer to avoid pointer false positives: 73 (cast(char[])buffer)[] = 0; 74 return true; 75 } 76 77 // First read. Determine if it's empty and put the char back. 78 auto mutableFP = (cast(File*) &file).getFP(); 79 const c = fgetc(mutableFP); 80 if (c == -1) 81 { 82 // Clean the buffer to avoid pointer false positives: 83 (cast(char[])buffer)[] = 0; 84 return true; 85 } 86 if (ungetc(c, mutableFP) != c) 87 { 88 assert(0, "Bug in cstdlib implementation"); 89 } 90 return false; 91 } 92 93 @property char[] front() @safe return scope 94 { 95 if (first_call) 96 { 97 popFront(); 98 first_call = false; 99 } 100 return line; 101 } 102 103 void popFront() @trusted scope 104 { 105 if (strBuffer.length == 0) 106 { 107 strBuffer = file.rawRead(buffer); 108 if (strBuffer.length == 0) 109 { 110 file.detach(); 111 line = null; 112 return; 113 } 114 } 115 116 // import std.string : indexOf; // TODO: array_algorithm indexOf 117 import nxt.array_algorithm : indexOf; 118 const pos = strBuffer.indexOf(this.separator); 119 if (pos != -1) 120 { 121 if (pos != 0 && strBuffer[pos-1] == '\r') 122 { 123 line = strBuffer[0 .. (pos-1)]; 124 } 125 else 126 { 127 line = strBuffer[0 .. pos]; 128 } 129 // Pop the line, skipping the terminator: 130 strBuffer = strBuffer[(pos+1) .. $]; 131 } 132 else 133 { 134 // More needs to be read here. Copy the tail of the buffer 135 // to the beginning, and try to read with the empty part of 136 // the buffer. 137 // If no buffer was left, extend the size of the buffer before 138 // reading. If the file has ended, then the line is the entire 139 // buffer. 140 141 if (strBuffer.ptr != buffer.ptr) 142 { 143 // Must use memmove because there might be overlap 144 memmove(buffer.ptr, strBuffer.ptr, 145 strBuffer.length * char.sizeof); 146 } 147 const spaceBegin = strBuffer.length; 148 if (strBuffer.length == buffer.length) 149 { 150 // Must extend the buffer to keep reading. 151 assumeSafeAppend(buffer); 152 buffer.length = buffer.length * 2; 153 } 154 const readPart = file.rawRead(buffer[spaceBegin .. $]); 155 if (readPart.length == 0) 156 { 157 // End of the file. Return whats in the buffer. 158 // The next popFront() will try to read again, and then 159 // mark empty condition. 160 if (spaceBegin != 0 && buffer[spaceBegin-1] == '\r') 161 { 162 line = buffer[0 .. spaceBegin-1]; 163 } 164 else 165 { 166 line = buffer[0 .. spaceBegin]; 167 } 168 strBuffer = null; 169 return; 170 } 171 strBuffer = buffer[0 .. spaceBegin + readPart.length]; 172 // Now that we have new data in strBuffer, we can go on. 173 // If a line isn't found, the buffer will be extended again to read more. 174 popFront(); 175 } 176 } 177 } 178 179 auto byLineFast(Terminator = char, 180 Char = char)(File f, 181 KeepTerminator keepTerminator = KeepTerminator.no, 182 string separator = "\n", 183 uint bufferSize = 4096) @safe // TODO: lookup preferred block type 184 { 185 return ByLineFast!(Char, Terminator)(f, keepTerminator, separator, bufferSize); 186 } 187 188 version(none): 189 190 version(linux) 191 unittest 192 { 193 import std.stdio: File, writeln; 194 import std.algorithm.searching: count; 195 import tempfs : tempFilePath; 196 import std.file : write; 197 198 const path = tempFilePath("x"); 199 200 writeln(path); 201 File(path, "wb").write("a\n"); 202 203 assert(File(path, "rb").byLineFast.count == 204 File(path, "rb").byLine.count); 205 } 206 207 unittest 208 { 209 import std.stdio: File, writeln; 210 import std.algorithm.searching: count; 211 212 const path = "/media/per/NORDLOW_2019-06-/Knowledge/DBpedia/latest/instance_types_en.ttl"; 213 214 import std.datetime: StopWatch; 215 216 double d1, d2; 217 218 { 219 StopWatch sw; 220 sw.start; 221 const c1 = File(path).byLine.count; 222 sw.stop; 223 d1 = sw.peek.msecs; 224 writeln("byLine: ", d1, "msecs"); 225 } 226 227 { 228 StopWatch sw; 229 sw.start; 230 const c2 = File(path).byLineFast.count; 231 sw.stop; 232 d2 = sw.peek.msecs; 233 writeln("byLineFast: ", d2, "msecs"); 234 } 235 236 writeln("Speed-Up: ", d1 / d2); 237 }