1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 2 "http://www.w3.org/TR/html4/strict.dtd"> 3 4<html> 5<head> 6 <title>Kaleidoscope: Tutorial Introduction and the Lexer</title> 7 <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> 8 <meta name="author" content="Chris Lattner"> 9 <link rel="stylesheet" href="../llvm.css" type="text/css"> 10</head> 11 12<body> 13 14<h1>Kaleidoscope: Tutorial Introduction and the Lexer</h1> 15 16<ul> 17<li><a href="index.html">Up to Tutorial Index</a></li> 18<li>Chapter 1 19 <ol> 20 <li><a href="#intro">Tutorial Introduction</a></li> 21 <li><a href="#language">The Basic Language</a></li> 22 <li><a href="#lexer">The Lexer</a></li> 23 </ol> 24</li> 25<li><a href="LangImpl2.html">Chapter 2</a>: Implementing a Parser and AST</li> 26</ul> 27 28<div class="doc_author"> 29 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a></p> 30</div> 31 32<!-- *********************************************************************** --> 33<h2><a name="intro">Tutorial Introduction</a></h2> 34<!-- *********************************************************************** --> 35 36<div> 37 38<p>Welcome to the "Implementing a language with LLVM" tutorial. This tutorial 39runs through the implementation of a simple language, showing how fun and 40easy it can be. This tutorial will get you up and started as well as help to 41build a framework you can extend to other languages. The code in this tutorial 42can also be used as a playground to hack on other LLVM specific things. 43</p> 44 45<p> 46The goal of this tutorial is to progressively unveil our language, describing 47how it is built up over time. This will let us cover a fairly broad range of 48language design and LLVM-specific usage issues, showing and explaining the code 49for it all along the way, without overwhelming you with tons of details up 50front.</p> 51 52<p>It is useful to point out ahead of time that this tutorial is really about 53teaching compiler techniques and LLVM specifically, <em>not</em> about teaching 54modern and sane software engineering principles. In practice, this means that 55we'll take a number of shortcuts to simplify the exposition. For example, the 56code leaks memory, uses global variables all over the place, doesn't use nice 57design patterns like <a 58href="http://en.wikipedia.org/wiki/Visitor_pattern">visitors</a>, etc... but it 59is very simple. If you dig in and use the code as a basis for future projects, 60fixing these deficiencies shouldn't be hard.</p> 61 62<p>I've tried to put this tutorial together in a way that makes chapters easy to 63skip over if you are already familiar with or are uninterested in the various 64pieces. The structure of the tutorial is: 65</p> 66 67<ul> 68<li><b><a href="#language">Chapter #1</a>: Introduction to the Kaleidoscope 69language, and the definition of its Lexer</b> - This shows where we are going 70and the basic functionality that we want it to do. In order to make this 71tutorial maximally understandable and hackable, we choose to implement 72everything in C++ instead of using lexer and parser generators. LLVM obviously 73works just fine with such tools, feel free to use one if you prefer.</li> 74<li><b><a href="LangImpl2.html">Chapter #2</a>: Implementing a Parser and 75AST</b> - With the lexer in place, we can talk about parsing techniques and 76basic AST construction. This tutorial describes recursive descent parsing and 77operator precedence parsing. Nothing in Chapters 1 or 2 is LLVM-specific, 78the code doesn't even link in LLVM at this point. :)</li> 79<li><b><a href="LangImpl3.html">Chapter #3</a>: Code generation to LLVM IR</b> - 80With the AST ready, we can show off how easy generation of LLVM IR really 81is.</li> 82<li><b><a href="LangImpl4.html">Chapter #4</a>: Adding JIT and Optimizer 83Support</b> - Because a lot of people are interested in using LLVM as a JIT, 84we'll dive right into it and show you the 3 lines it takes to add JIT support. 85LLVM is also useful in many other ways, but this is one simple and "sexy" way 86to shows off its power. :)</li> 87<li><b><a href="LangImpl5.html">Chapter #5</a>: Extending the Language: Control 88Flow</b> - With the language up and running, we show how to extend it with 89control flow operations (if/then/else and a 'for' loop). This gives us a chance 90to talk about simple SSA construction and control flow.</li> 91<li><b><a href="LangImpl6.html">Chapter #6</a>: Extending the Language: 92User-defined Operators</b> - This is a silly but fun chapter that talks about 93extending the language to let the user program define their own arbitrary 94unary and binary operators (with assignable precedence!). This lets us build a 95significant piece of the "language" as library routines.</li> 96<li><b><a href="LangImpl7.html">Chapter #7</a>: Extending the Language: Mutable 97Variables</b> - This chapter talks about adding user-defined local variables 98along with an assignment operator. The interesting part about this is how 99easy and trivial it is to construct SSA form in LLVM: no, LLVM does <em>not</em> 100require your front-end to construct SSA form!</li> 101<li><b><a href="LangImpl8.html">Chapter #8</a>: Conclusion and other useful LLVM 102tidbits</b> - This chapter wraps up the series by talking about potential 103ways to extend the language, but also includes a bunch of pointers to info about 104"special topics" like adding garbage collection support, exceptions, debugging, 105support for "spaghetti stacks", and a bunch of other tips and tricks.</li> 106 107</ul> 108 109<p>By the end of the tutorial, we'll have written a bit less than 700 lines of 110non-comment, non-blank, lines of code. With this small amount of code, we'll 111have built up a very reasonable compiler for a non-trivial language including 112a hand-written lexer, parser, AST, as well as code generation support with a JIT 113compiler. While other systems may have interesting "hello world" tutorials, 114I think the breadth of this tutorial is a great testament to the strengths of 115LLVM and why you should consider it if you're interested in language or compiler 116design.</p> 117 118<p>A note about this tutorial: we expect you to extend the language and play 119with it on your own. Take the code and go crazy hacking away at it, compilers 120don't need to be scary creatures - it can be a lot of fun to play with 121languages!</p> 122 123</div> 124 125<!-- *********************************************************************** --> 126<h2><a name="language">The Basic Language</a></h2> 127<!-- *********************************************************************** --> 128 129<div> 130 131<p>This tutorial will be illustrated with a toy language that we'll call 132"<a href="http://en.wikipedia.org/wiki/Kaleidoscope">Kaleidoscope</a>" (derived 133from "meaning beautiful, form, and view"). 134Kaleidoscope is a procedural language that allows you to define functions, use 135conditionals, math, etc. Over the course of the tutorial, we'll extend 136Kaleidoscope to support the if/then/else construct, a for loop, user defined 137operators, JIT compilation with a simple command line interface, etc.</p> 138 139<p>Because we want to keep things simple, the only datatype in Kaleidoscope is a 14064-bit floating point type (aka 'double' in C parlance). As such, all values 141are implicitly double precision and the language doesn't require type 142declarations. This gives the language a very nice and simple syntax. For 143example, the following simple example computes <a 144href="http://en.wikipedia.org/wiki/Fibonacci_number">Fibonacci numbers:</a></p> 145 146<div class="doc_code"> 147<pre> 148# Compute the x'th fibonacci number. 149def fib(x) 150 if x < 3 then 151 1 152 else 153 fib(x-1)+fib(x-2) 154 155# This expression will compute the 40th number. 156fib(40) 157</pre> 158</div> 159 160<p>We also allow Kaleidoscope to call into standard library functions (the LLVM 161JIT makes this completely trivial). This means that you can use the 'extern' 162keyword to define a function before you use it (this is also useful for mutually 163recursive functions). For example:</p> 164 165<div class="doc_code"> 166<pre> 167extern sin(arg); 168extern cos(arg); 169extern atan2(arg1 arg2); 170 171atan2(sin(.4), cos(42)) 172</pre> 173</div> 174 175<p>A more interesting example is included in Chapter 6 where we write a little 176Kaleidoscope application that <a href="LangImpl6.html#example">displays 177a Mandelbrot Set</a> at various levels of magnification.</p> 178 179<p>Lets dive into the implementation of this language!</p> 180 181</div> 182 183<!-- *********************************************************************** --> 184<h2><a name="lexer">The Lexer</a></h2> 185<!-- *********************************************************************** --> 186 187<div> 188 189<p>When it comes to implementing a language, the first thing needed is 190the ability to process a text file and recognize what it says. The traditional 191way to do this is to use a "<a 192href="http://en.wikipedia.org/wiki/Lexical_analysis">lexer</a>" (aka 'scanner') 193to break the input up into "tokens". Each token returned by the lexer includes 194a token code and potentially some metadata (e.g. the numeric value of a number). 195First, we define the possibilities: 196</p> 197 198<div class="doc_code"> 199<pre> 200// The lexer returns tokens [0-255] if it is an unknown character, otherwise one 201// of these for known things. 202enum Token { 203 tok_eof = -1, 204 205 // commands 206 tok_def = -2, tok_extern = -3, 207 208 // primary 209 tok_identifier = -4, tok_number = -5, 210}; 211 212static std::string IdentifierStr; // Filled in if tok_identifier 213static double NumVal; // Filled in if tok_number 214</pre> 215</div> 216 217<p>Each token returned by our lexer will either be one of the Token enum values 218or it will be an 'unknown' character like '+', which is returned as its ASCII 219value. If the current token is an identifier, the <tt>IdentifierStr</tt> 220global variable holds the name of the identifier. If the current token is a 221numeric literal (like 1.0), <tt>NumVal</tt> holds its value. Note that we use 222global variables for simplicity, this is not the best choice for a real language 223implementation :). 224</p> 225 226<p>The actual implementation of the lexer is a single function named 227<tt>gettok</tt>. The <tt>gettok</tt> function is called to return the next token 228from standard input. Its definition starts as:</p> 229 230<div class="doc_code"> 231<pre> 232/// gettok - Return the next token from standard input. 233static int gettok() { 234 static int LastChar = ' '; 235 236 // Skip any whitespace. 237 while (isspace(LastChar)) 238 LastChar = getchar(); 239</pre> 240</div> 241 242<p> 243<tt>gettok</tt> works by calling the C <tt>getchar()</tt> function to read 244characters one at a time from standard input. It eats them as it recognizes 245them and stores the last character read, but not processed, in LastChar. The 246first thing that it has to do is ignore whitespace between tokens. This is 247accomplished with the loop above.</p> 248 249<p>The next thing <tt>gettok</tt> needs to do is recognize identifiers and 250specific keywords like "def". Kaleidoscope does this with this simple loop:</p> 251 252<div class="doc_code"> 253<pre> 254 if (isalpha(LastChar)) { // identifier: [a-zA-Z][a-zA-Z0-9]* 255 IdentifierStr = LastChar; 256 while (isalnum((LastChar = getchar()))) 257 IdentifierStr += LastChar; 258 259 if (IdentifierStr == "def") return tok_def; 260 if (IdentifierStr == "extern") return tok_extern; 261 return tok_identifier; 262 } 263</pre> 264</div> 265 266<p>Note that this code sets the '<tt>IdentifierStr</tt>' global whenever it 267lexes an identifier. Also, since language keywords are matched by the same 268loop, we handle them here inline. Numeric values are similar:</p> 269 270<div class="doc_code"> 271<pre> 272 if (isdigit(LastChar) || LastChar == '.') { // Number: [0-9.]+ 273 std::string NumStr; 274 do { 275 NumStr += LastChar; 276 LastChar = getchar(); 277 } while (isdigit(LastChar) || LastChar == '.'); 278 279 NumVal = strtod(NumStr.c_str(), 0); 280 return tok_number; 281 } 282</pre> 283</div> 284 285<p>This is all pretty straight-forward code for processing input. When reading 286a numeric value from input, we use the C <tt>strtod</tt> function to convert it 287to a numeric value that we store in <tt>NumVal</tt>. Note that this isn't doing 288sufficient error checking: it will incorrectly read "1.23.45.67" and handle it as 289if you typed in "1.23". Feel free to extend it :). Next we handle comments: 290</p> 291 292<div class="doc_code"> 293<pre> 294 if (LastChar == '#') { 295 // Comment until end of line. 296 do LastChar = getchar(); 297 while (LastChar != EOF && LastChar != '\n' && LastChar != '\r'); 298 299 if (LastChar != EOF) 300 return gettok(); 301 } 302</pre> 303</div> 304 305<p>We handle comments by skipping to the end of the line and then return the 306next token. Finally, if the input doesn't match one of the above cases, it is 307either an operator character like '+' or the end of the file. These are handled 308with this code:</p> 309 310<div class="doc_code"> 311<pre> 312 // Check for end of file. Don't eat the EOF. 313 if (LastChar == EOF) 314 return tok_eof; 315 316 // Otherwise, just return the character as its ascii value. 317 int ThisChar = LastChar; 318 LastChar = getchar(); 319 return ThisChar; 320} 321</pre> 322</div> 323 324<p>With this, we have the complete lexer for the basic Kaleidoscope language 325(the <a href="LangImpl2.html#code">full code listing</a> for the Lexer is 326available in the <a href="LangImpl2.html">next chapter</a> of the tutorial). 327Next we'll <a href="LangImpl2.html">build a simple parser that uses this to 328build an Abstract Syntax Tree</a>. When we have that, we'll include a driver 329so that you can use the lexer and parser together. 330</p> 331 332<a href="LangImpl2.html">Next: Implementing a Parser and AST</a> 333</div> 334 335<!-- *********************************************************************** --> 336<hr> 337<address> 338 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img 339 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a> 340 <a href="http://validator.w3.org/check/referer"><img 341 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a> 342 343 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br> 344 <a href="http://llvm.org/">The LLVM Compiler Infrastructure</a><br> 345 Last modified: $Date: 2011-04-22 20:30:22 -0400 (Fri, 22 Apr 2011) $ 346</address> 347</body> 348</html> 349