Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / clang / 336f26be05c30c0c65b8b518d68f04c29b43b3d2 / . / docs / InternalsManual.html
blob: 89111dac470a474e67df5bc59d8e55bb897fcbb3 [file] [log] [blame]
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1<html>
2<head>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]3<title>"Clang" CFE Internals Manual</title>
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]4<link type="text/css" rel="stylesheet" href="../menu.css" />
5<link type="text/css" rel="stylesheet" href="../content.css" />
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]6<style type="text/css">
7td {
8 vertical-align: top;
9}
10</style>
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]11</head>
12<body>
13
14<!--#include virtual="../menu.html.incl"-->
15
16<div id="content">
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]17
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]18<h1>"Clang" CFE Internals Manual</h1>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]19
20<ul>
21<li><a href="#intro">Introduction</a></li>
22<li><a href="#libsystem">LLVM System and Support Libraries</a></li>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]23<li><a href="#libbasic">The Clang 'Basic' Library</a>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]24 <ul>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]25 <li><a href="#Diagnostics">The Diagnostics Subsystem</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]26 <li><a href="#SourceLocation">The SourceLocation and SourceManager
27 classes</a></li>
28 </ul>
29</li>
30<li><a href="#liblex">The Lexer and Preprocessor Library</a>
31 <ul>
32 <li><a href="#Token">The Token class</a></li>
33 <li><a href="#Lexer">The Lexer class</a></li>
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]34 <li><a href="#AnnotationToken">Annotation Tokens</a></li>
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]35 <li><a href="#TokenLexer">The TokenLexer class</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]36 <li><a href="#MultipleIncludeOpt">The MultipleIncludeOpt class</a></li>
37 </ul>
38</li>
39<li><a href="#libparse">The Parser Library</a>
40 <ul>
41 </ul>
42</li>
43<li><a href="#libast">The AST Library</a>
44 <ul>
45 <li><a href="#Type">The Type class and its subclasses</a></li>
46 <li><a href="#QualType">The QualType class</a></li>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]47 <li><a href="#DeclarationName">Declaration names</a></li>
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]48 <li><a href="#DeclContext">Declaration contexts</a>
49 <ul>
50 <li><a href="#Redeclarations">Redeclarations and Overloads</a></li>
51 <li><a href="#LexicalAndSemanticContexts">Lexical and Semantic
52 Contexts</a></li>
53 <li><a href="#TransparentContexts">Transparent Declaration Contexts</a></li>
54 <li><a href="#MultiDeclContext">Multiply-Defined Declaration Contexts</a></li>
55 </ul>
56 </li>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]57 <li><a href="#CFG">The CFG class</a></li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]58 <li><a href="#Constants">Constant Folding in the Clang AST</a></li>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]59 </ul>
60</li>
61</ul>
62
63
64<!-- ======================================================================= -->
65<h2 id="intro">Introduction</h2>
66<!-- ======================================================================= -->
67
68<p>This document describes some of the more important APIs and internal design
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]69decisions made in the Clang C front-end. The purpose of this document is to
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]70both capture some of this high level information and also describe some of the
71design decisions behind it. This is meant for people interested in hacking on
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]72Clang, not for end-users. The description below is categorized by
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]73libraries, and does not describe any of the clients of the libraries.</p>
74
75<!-- ======================================================================= -->
76<h2 id="libsystem">LLVM System and Support Libraries</h2>
77<!-- ======================================================================= -->
78
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]79<p>The LLVM libsystem library provides the basic Clang system abstraction layer,
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]80which is used for file system access. The LLVM libsupport library provides many
81underlying libraries and <a
82href="http://llvm.org/docs/ProgrammersManual.html">data-structures</a>,
83 including command line option
84processing and various containers.</p>
85
86<!-- ======================================================================= -->
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]87<h2 id="libbasic">The Clang 'Basic' Library</h2>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]88<!-- ======================================================================= -->
89
90<p>This library certainly needs a better name. The 'basic' library contains a
91number of low-level utilities for tracking and manipulating source buffers,
92locations within the source buffers, diagnostics, tokens, target abstraction,
93and information about the subset of the language being compiled for.</p>
94
95<p>Part of this infrastructure is specific to C (such as the TargetInfo class),
96other parts could be reused for other non-C-based languages (SourceLocation,
97SourceManager, Diagnostics, FileManager). When and if there is future demand
98we can figure out if it makes sense to introduce a new library, move the general
99classes somewhere else, or introduce some other solution.</p>
100
101<p>We describe the roles of these classes in order of their dependencies.</p>
102
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]103
104<!-- ======================================================================= -->
105<h3 id="Diagnostics">The Diagnostics Subsystem</h3>
106<!-- ======================================================================= -->
107
108<p>The Clang Diagnostics subsystem is an important part of how the compiler
109communicates with the human. Diagnostics are the warnings and errors produced
110when the code is incorrect or dubious. In Clang, each diagnostic produced has
111(at the minimum) a unique ID, a <a href="#SourceLocation">SourceLocation</a> to
112"put the caret", an English translation associated with it, and a severity (e.g.
113<tt>WARNING</tt> or <tt>ERROR</tt>). They can also optionally include a number
114of arguments to the dianostic (which fill in "%0"'s in the string) as well as a
115number of source ranges that related to the diagnostic.</p>
116
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]117<p>In this section, we'll be giving examples produced by the Clang command line
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]118driver, but diagnostics can be <a href="#DiagnosticClient">rendered in many
119different ways</a> depending on how the DiagnosticClient interface is
120implemented. A representative example of a diagonstic is:</p>
121
122<pre>
123t.c:38:15: error: invalid operands to binary expression ('int *' and '_Complex float')
124 <font color="darkgreen">P = (P-42) + Gamma*4;</font>
125 <font color="blue">~~~~~~ ^ ~~~~~~~</font>
126</pre>
127
128<p>In this example, you can see the English translation, the severity (error),
129you can see the source location (the caret ("^") and file/line/column info),
130the source ranges "~~~~", arguments to the diagnostic ("int*" and "_Complex
131float"). You'll have to believe me that there is a unique ID backing the
132diagnostic :).</p>
133
134<p>Getting all of this to happen has several steps and involves many moving
135pieces, this section describes them and talks about best practices when adding
136a new diagnostic.</p>
137
138<!-- ============================ -->
139<h4>The DiagnosticKinds.def file</h4>
140<!-- ============================ -->
141
142<p>Diagnostics are created by adding an entry to the <tt><a
143href="http://llvm.org/svn/llvm-project/cfe/trunk/include/clang/Basic/DiagnosticKinds.def"
144>DiagnosticKinds.def</a></tt> file. This file encodes the unique ID of the
145diagnostic (as an enum, the first argument), the severity of the diagnostic
146(second argument) and the English translation + format string.</p>
147
148<p>There is little sanity with the naming of the unique ID's right now. Some
149start with err_, warn_, ext_ to encode the severity into the name. Since the
150enum is referenced in the C++ code that produces the diagnostic, it is somewhat
151useful for it to be reasonably short.</p>
152
153<p>The severity of the diagnostic comes from the set {<tt>NOTE</tt>,
154<tt>WARNING</tt>, <tt>EXTENSION</tt>, <tt>EXTWARN</tt>, <tt>ERROR</tt>}. The
155<tt>ERROR</tt> severity is used for diagnostics indicating the program is never
156acceptable under any circumstances. When an error is emitted, the AST for the
157input code may not be fully built. The <tt>EXTENSION</tt> and <tt>EXTWARN</tt>
158severities are used for extensions to the language that Clang accepts. This
159means that Clang fully understands and can represent them in the AST, but we
160produce diagnostics to tell the user their code is non-portable. The difference
161is that the former are ignored by default, and the later warn by default. The
162<tt>WARNING</tt> severity is used for constructs that are valid in the currently
163selected source language but that are dubious in some way. The <tt>NOTE</tt>
164level is used to staple more information onto a previous diagnostics.</p>
165
166<p>These <em>severities</em> are mapped into a smaller set (the
167Diagnostic::Level enum, {<tt>Ignored</tt>, <tt>Note</tt>, <tt>Warning</tt>,
Chris Lattner0aad2972009-02-05 22:49:08 +0000[diff] [blame]168<tt>Error</tt>, <tt>Fatal</tt> }) of output <em>levels</em> by the diagnostics
169subsystem based
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]170on various configuration options. For example, if the user specifies
171<tt>-pedantic</tt>, <tt>EXTENSION</tt> maps to <tt>Warning</tt>, if they specify
172<tt>-pedantic-errors</tt>, it turns into <tt>Error</tt>. Clang also internally
173supports a fully fine grained mapping mechanism that allows you to map any
174diagnostic that doesn't have <tt>ERRROR</tt> severity to any output level that
175you want. This is used to implement options like <tt>-Wunused_macros</tt>,
Chris Lattner0aad2972009-02-05 22:49:08 +0000[diff] [blame]176<tt>-Wundef</tt> etc. Fatal errors are considered so severe that diagnostics
177that occur after them are supressed as "almost certainly useless".</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]178
179<!-- ================= -->
180<h4>The Format String</h4>
181<!-- ================= -->
182
183<p>The format string for the diagnostic is very simple, but it has some power.
184It takes the form of a string in English with markers that indicate where and
185how arguments to the diagnostic are inserted and formatted. For example, here
186are some simple format strings:</p>
187
188<pre>
189 "binary integer literals are an extension"
190 "format string contains '\\0' within the string body"
191 "more '<b>%%</b>' conversions than data arguments"
Chris Lattner545b3682008-11-23 20:27:13 +0000[diff] [blame]192 "invalid operands to binary expression (<b>%0</b> and <b>%1</b>)"
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]193 "overloaded '<b>%0</b>' must be a <b>%select{unary|binary|unary or binary}2</b> operator"
194 " (has <b>%1</b> parameter<b>%s1</b>)"
195</pre>
196
197<p>These examples show some important points of format strings. You can use any
198 plain ASCII character in the diagnostic string except "%" without a problem,
199 but these are C strings, so you have to use and be aware of all the C escape
200 sequences (as in the second example). If you want to produce a "%" in the
201 output, use the "%%" escape sequence, like the third diagnostic. Finally,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]202 Clang uses the "%...[digit]" sequences to specify where and how arguments to
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]203 the diagnostic are formatted.</p>
204
205<p>Arguments to the diagnostic are numbered according to how they are specified
206 by the C++ code that <a href="#producingdiag">produces them</a>, and are
207 referenced by <tt>%0</tt> .. <tt>%9</tt>. If you have more than 10 arguments
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]208 to your diagnostic, you are doing something wrong :). Unlike printf, there
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]209 is no requirement that arguments to the diagnostic end up in the output in
210 the same order as they are specified, you could have a format string with
211 <tt>"%1 %0"</tt> that swaps them, for example. The text in between the
212 percent and digit are formatting instructions. If there are no instructions,
213 the argument is just turned into a string and substituted in.</p>
214
215<p>Here are some "best practices" for writing the English format string:</p>
216
217<ul>
218<li>Keep the string short. It should ideally fit in the 80 column limit of the
219 <tt>DiagnosticKinds.def</tt> file. This avoids the diagnostic wrapping when
220 printed, and forces you to think about the important point you are conveying
221 with the diagnostic.</li>
222<li>Take advantage of location information. The user will be able to see the
223 line and location of the caret, so you don't need to tell them that the
224 problem is with the 4th argument to the function: just point to it.</li>
225<li>Do not capitalize the diagnostic string, and do not end it with a
226 period.</li>
227<li>If you need to quote something in the diagnostic string, use single
228 quotes.</li>
229</ul>
230
231<p>Diagnostics should never take random English strings as arguments: you
232shouldn't use <tt>"you have a problem with %0"</tt> and pass in things like
233<tt>"your argument"</tt> or <tt>"your return value"</tt> as arguments. Doing
234this prevents <a href="translation">translating</a> the Clang diagnostics to
235other languages (because they'll get random English words in their otherwise
236localized diagnostic). The exceptions to this are C/C++ language keywords
237(e.g. auto, const, mutable, etc) and C/C++ operators (<tt>/=</tt>). Note
238that things like "pointer" and "reference" are not keywords. On the other
239hand, you <em>can</em> include anything that comes from the user's source code,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]240including variable names, types, labels, etc. The 'select' format can be
241used to achieve this sort of thing in a localizable way, see below.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]242
243<!-- ==================================== -->
244<h4>Formatting a Diagnostic Argument</a></h4>
245<!-- ==================================== -->
246
247<p>Arguments to diagnostics are fully typed internally, and come from a couple
248different classes: integers, types, names, and random strings. Depending on
249the class of the argument, it can be optionally formatted in different ways.
250This gives the DiagnosticClient information about what the argument means
251without requiring it to use a specific presentation (consider this MVC for
252Clang :).</p>
253
254<p>Here are the different diagnostic argument formats currently supported by
255Clang:</p>
256
257<table>
258<tr><td colspan="2"><b>"s" format</b></td></tr>
259<tr><td>Example:</td><td><tt>"requires %1 parameter%s1"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]260<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]261<tr><td>Description:</td><td>This is a simple formatter for integers that is
262 useful when producing English diagnostics. When the integer is 1, it prints
263 as nothing. When the integer is not 1, it prints as "s". This allows some
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]264 simple grammatical forms to be to be handled correctly, and eliminates the
265 need to use gross things like <tt>"requires %1 parameter(s)"</tt>.</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]266
267<tr><td colspan="2"><b>"select" format</b></td></tr>
268<tr><td>Example:</td><td><tt>"must be a %select{unary|binary|unary or binary}2
269 operator"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]270<tr><td>Class:</td><td>Integers</td></tr>
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]271<tr><td>Description:</td><td>This format specifier is used to merge multiple
272 related diagnostics together into one common one, without requiring the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]273 difference to be specified as an English string argument. Instead of
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]274 specifying the string, the diagnostic gets an integer argument and the
275 format string selects the numbered option. In this case, the "%2" value
276 must be an integer in the range [0..2]. If it is 0, it prints 'unary', if
277 it is 1 it prints 'binary' if it is 2, it prints 'unary or binary'. This
278 allows other language translations to substitute reasonable words (or entire
279 phrases) based on the semantics of the diagnostic instead of having to do
280 things textually.</td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]281
282<tr><td colspan="2"><b>"plural" format</b></td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]283<tr><td>Example:</td><td><tt>"you have %1 %plural{1:mouse|:mice}1 connected to
284 your computer"</tt></td></tr>
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]285<tr><td>Class:</td><td>Integers</td></tr>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]286<tr><td>Description:</td><td><p>This is a formatter for complex plural forms.
287 It is designed to handle even the requirements of languages with very
288 complex plural forms, as many Baltic languages have. The argument consists
289 of a series of expression/form pairs, separated by ':', where the first form
290 whose expression evaluates to true is the result of the modifier.</p>
291 <p>An expression can be empty, in which case it is always true. See the
292 example at the top. Otherwise, it is a series of one or more numeric
293 conditions, separated by ','. If any condition matches, the expression
294 matches. Each numeric condition can take one of three forms.</p>
295 <ul>
296 <li>number: A simple decimal number matches if the argument is the same
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]297 as the number. Example: <tt>"%plural{1:mouse|:mice}4"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]298 <li>range: A range in square brackets matches if the argument is within
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]299 the range. Then range is inclusive on both ends. Example:
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]300 <tt>"%plural{0:none|1:one|[2,5]:some|:many}2"</tt></li>
301 <li>modulo: A modulo operator is followed by a number, and
302 equals sign and either a number or a range. The tests are the
303 same as for plain
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]304 numbers and ranges, but the argument is taken modulo the number first.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]305 Example: <tt>"%plural{%100=0:even hundred|%100=[1,50]:lower half|:everything
306 else}1"</tt></li>
Sebastian Redl68168562008-11-22 22:16:45 +0000[diff] [blame]307 </ul>
308 <p>The parser is very unforgiving. A syntax error, even whitespace, will
309 abort, as will a failure to match the argument against any
310 expression.</p></td></tr>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]311
Chris Lattner077bf5e2008-11-24 03:33:13 +0000[diff] [blame]312<tr><td colspan="2"><b>"objcclass" format</b></td></tr>
313<tr><td>Example:</td><td><tt>"method %objcclass0 not found"</tt></td></tr>
314<tr><td>Class:</td><td>DeclarationName</td></tr>
315<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
316 DeclarationName corresponds to an Objective-C class method selector. As
317 such, it prints the selector with a leading '+'.</p></td></tr>
318
319<tr><td colspan="2"><b>"objcinstance" format</b></td></tr>
320<tr><td>Example:</td><td><tt>"method %objcinstance0 not found"</tt></td></tr>
321<tr><td>Class:</td><td>DeclarationName</td></tr>
322<tr><td>Description:</td><td><p>This is a simple formatter that indicates the
323 DeclarationName corresponds to an Objective-C instance method selector. As
324 such, it prints the selector with a leading '-'.</p></td></tr>
325
Douglas Gregor47b9a1c2009-02-04 17:27:36 +0000[diff] [blame]326<tr><td colspan="2"><b>"q" format</b></td></tr>
327<tr><td>Example:</td><td><tt>"candidate found by name lookup is %q0"</tt></td></tr>
328<tr><td>Class:</td><td>NamedDecl*</td></tr>
329<tr><td>Description</td><td><p>This formatter indicates that the fully-qualified name of the declaration should be printed, e.g., "std::vector" rather than "vector".</p></td></tr>
330
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]331</table>
332
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]333<p>It is really easy to add format specifiers to the Clang diagnostics system,
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]334but they should be discussed before they are added. If you are creating a lot
335of repetitive diagnostics and/or have an idea for a useful formatter, please
336bring it up on the cfe-dev mailing list.</p>
Chris Lattnercc543342008-11-22 23:50:47 +0000[diff] [blame]337
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]338<!-- ===================================================== -->
339<h4><a name="#producingdiag">Producing the Diagnostic</a></h4>
340<!-- ===================================================== -->
341
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]342<p>Now that you've created the diagnostic in the DiagnosticKinds.def file, you
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]343need to write the code that detects the condition in question and emits the
344new diagnostic. Various components of Clang (e.g. the preprocessor, Sema,
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]345etc) provide a helper function named "Diag". It creates a diagnostic and
346accepts the arguments, ranges, and other information that goes along with
347it.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]348
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]349<p>For example, the binary expression error comes from code like this:</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]350
351<pre>
352 if (various things that are bad)
353 Diag(Loc, diag::err_typecheck_invalid_operands)
354 &lt;&lt; lex-&gt;getType() &lt;&lt; rex-&gt;getType()
355 &lt;&lt; lex-&gt;getSourceRange() &lt;&lt; rex-&gt;getSourceRange();
356</pre>
357
358<p>This shows that use of the Diag method: they take a location (a <a
359href="#SourceLocation">SourceLocation</a> object) and a diagnostic enum value
360(which matches the name from DiagnosticKinds.def). If the diagnostic takes
361arguments, they are specified with the &lt;&lt; operator: the first argument
362becomes %0, the second becomes %1, etc. The diagnostic interface allows you to
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]363specify arguments of many different types, including <tt>int</tt> and
364<tt>unsigned</tt> for integer arguments, <tt>const char*</tt> and
365<tt>std::string</tt> for string arguments, <tt>DeclarationName</tt> and
366<tt>const IdentifierInfo*</tt> for names, <tt>QualType</tt> for types, etc.
367SourceRanges are also specified with the &lt;&lt; operator, but do not have a
368specific ordering requirement.</p>
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]369
370<p>As you can see, adding and producing a diagnostic is pretty straightforward.
371The hard part is deciding exactly what you need to say to help the user, picking
372a suitable wording, and providing the information needed to format it correctly.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]373The good news is that the call site that issues a diagnostic should be
374completely independent of how the diagnostic is formatted and in what language
375it is rendered.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]376</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]377
378<!-- ============================================================= -->
379<h4><a name="DiagnosticClient">The DiagnosticClient Interface</a></h4>
380<!-- ============================================================= -->
381
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]382<p>Once code generates a diagnostic with all of the arguments and the rest of
383the relevant information, Clang needs to know what to do with it. As previously
384mentioned, the diagnostic machinery goes through some filtering to map a
385severity onto a diagnostic level, then (assuming the diagnostic is not mapped to
386"<tt>Ignore</tt>") it invokes an object that implements the DiagnosticClient
387interface with the information.</p>
388
389<p>It is possible to implement this interface in many different ways. For
390example, the normal Clang DiagnosticClient (named 'TextDiagnosticPrinter') turns
391the arguments into strings (according to the various formatting rules), prints
392out the file/line/column information and the string, then prints out the line of
393code, the source ranges, and the caret. However, this behavior isn't required.
394</p>
395
396<p>Another implementation of the DiagnosticClient interface is the
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]397'TextDiagnosticBuffer' class, which is used when Clang is in -verify mode.
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]398Instead of formatting and printing out the diagnostics, this implementation just
399captures and remembers the diagnostics as they fly by. Then -verify compares
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]400the list of produced diagnostics to the list of expected ones. If they disagree,
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]401it prints out its own output.
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]402</p>
403
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]404<p>There are many other possible implementations of this interface, and this is
405why we prefer diagnostics to pass down rich structured information in arguments.
406For example, an HTML output might want declaration names be linkified to where
407they come from in the source. Another example is that a GUI might let you click
408on typedefs to expand them. This application would want to pass significantly
409more information about types through to the GUI than a simple flat string. The
410interface allows this to happen.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]411
412<!-- ====================================================== -->
413<h4><a name="translation">Adding Translations to Clang</a></h4>
414<!-- ====================================================== -->
415
Chris Lattner627b7052008-11-23 00:28:33 +0000[diff] [blame]416<p>Not possible yet! Diagnostic strings should be written in UTF-8, the client
Chris Lattnerfd408ea2008-11-23 00:42:53 +0000[diff] [blame]417can translate to the relevant code page if needed. Each translation completely
418replaces the format string for the diagnostic.</p>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]419
420
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]421<!-- ======================================================================= -->
422<h3 id="SourceLocation">The SourceLocation and SourceManager classes</h3>
423<!-- ======================================================================= -->
424
425<p>Strangely enough, the SourceLocation class represents a location within the
426source code of the program. Important design points include:</p>
427
428<ol>
429<li>sizeof(SourceLocation) must be extremely small, as these are embedded into
430 many AST nodes and are passed around often. Currently it is 32 bits.</li>
431<li>SourceLocation must be a simple value object that can be efficiently
432 copied.</li>
433<li>We should be able to represent a source location for any byte of any input
434 file. This includes in the middle of tokens, in whitespace, in trigraphs,
435 etc.</li>
436<li>A SourceLocation must encode the current #include stack that was active when
437 the location was processed. For example, if the location corresponds to a
438 token, it should contain the set of #includes active when the token was
439 lexed. This allows us to print the #include stack for a diagnostic.</li>
440<li>SourceLocation must be able to describe macro expansions, capturing both
441 the ultimate instantiation point and the source of the original character
442 data.</li>
443</ol>
444
445<p>In practice, the SourceLocation works together with the SourceManager class
Chris Lattner18376dd2009-01-16 07:00:50 +0000[diff] [blame]446to encode two pieces of information about a location: it's spelling location
Chris Lattner88054de2009-01-16 07:15:35 +0000[diff] [blame]447and it's instantiation location. For most tokens, these will be the same. However,
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]448for a macro expansion (or tokens that came from a _Pragma directive) these will
449describe the location of the characters corresponding to the token and the
450location where the token was used (i.e. the macro instantiation point or the
451location of the _Pragma itself).</p>
452
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]453<p>For efficiency, we only track one level of macro instantiations: if a token was
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]454produced by multiple instantiations, we only track the source and ultimate
455destination. Though we could track the intermediate instantiation points, this
456would require extra bookkeeping and no known client would benefit substantially
457from this.</p>
458
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]459<p>The Clang front-end inherently depends on the location of a token being
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]460tracked correctly. If it is ever incorrect, the front-end may get confused and
461die. The reason for this is that the notion of the 'spelling' of a Token in
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]462Clang depends on being able to find the original input characters for the token.
Chris Lattner18376dd2009-01-16 07:00:50 +0000[diff] [blame]463This concept maps directly to the "spelling location" for the token.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]464
465<!-- ======================================================================= -->
466<h2 id="liblex">The Lexer and Preprocessor Library</h2>
467<!-- ======================================================================= -->
468
469<p>The Lexer library contains several tightly-connected classes that are involved
470with the nasty process of lexing and preprocessing C source code. The main
471interface to this library for outside clients is the large <a
472href="#Preprocessor">Preprocessor</a> class.
473It contains the various pieces of state that are required to coherently read
474tokens out of a translation unit.</p>
475
476<p>The core interface to the Preprocessor object (once it is set up) is the
477Preprocessor::Lex method, which returns the next <a href="#Token">Token</a> from
478the preprocessor stream. There are two types of token providers that the
479preprocessor is capable of reading from: a buffer lexer (provided by the <a
480href="#Lexer">Lexer</a> class) and a buffered token stream (provided by the <a
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]481href="#TokenLexer">TokenLexer</a> class).
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]482
483
484<!-- ======================================================================= -->
485<h3 id="Token">The Token class</h3>
486<!-- ======================================================================= -->
487
488<p>The Token class is used to represent a single lexed token. Tokens are
489intended to be used by the lexer/preprocess and parser libraries, but are not
490intended to live beyond them (for example, they should not live in the ASTs).<p>
491
492<p>Tokens most often live on the stack (or some other location that is efficient
493to access) as the parser is running, but occasionally do get buffered up. For
494example, macro definitions are stored as a series of tokens, and the C++
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]495front-end periodically needs to buffer tokens up for tentative parsing and
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]496various pieces of look-ahead. As such, the size of a Token matter. On a 32-bit
497system, sizeof(Token) is currently 16 bytes.</p>
498
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]499<p>Tokens occur in two forms: "<a href="#AnnotationToken">Annotation
500Tokens</a>" and normal tokens. Normal tokens are those returned by the lexer,
501annotation tokens represent semantic information and are produced by the parser,
502replacing normal tokens in the token stream. Normal tokens contain the
503following information:</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]504
505<ul>
506<li><b>A SourceLocation</b> - This indicates the location of the start of the
507token.</li>
508
509<li><b>A length</b> - This stores the length of the token as stored in the
510SourceBuffer. For tokens that include them, this length includes trigraphs and
511escaped newlines which are ignored by later phases of the compiler. By pointing
512into the original source buffer, it is always possible to get the original
513spelling of a token completely accurately.</li>
514
515<li><b>IdentifierInfo</b> - If a token takes the form of an identifier, and if
516identifier lookup was enabled when the token was lexed (e.g. the lexer was not
517reading in 'raw' mode) this contains a pointer to the unique hash value for the
518identifier. Because the lookup happens before keyword identification, this
519field is set even for language keywords like 'for'.</li>
520
521<li><b>TokenKind</b> - This indicates the kind of token as classified by the
522lexer. This includes things like <tt>tok::starequal</tt> (for the "*="
523operator), <tt>tok::ampamp</tt> for the "&amp;&amp;" token, and keyword values
524(e.g. <tt>tok::kw_for</tt>) for identifiers that correspond to keywords. Note
525that some tokens can be spelled multiple ways. For example, C++ supports
526"operator keywords", where things like "and" are treated exactly like the
527"&amp;&amp;" operator. In these cases, the kind value is set to
528<tt>tok::ampamp</tt>, which is good for the parser, which doesn't have to
529consider both forms. For something that cares about which form is used (e.g.
530the preprocessor 'stringize' operator) the spelling indicates the original
531form.</li>
532
533<li><b>Flags</b> - There are currently four flags tracked by the
534lexer/preprocessor system on a per-token basis:
535
536 <ol>
537 <li><b>StartOfLine</b> - This was the first token that occurred on its input
538 source line.</li>
539 <li><b>LeadingSpace</b> - There was a space character either immediately
540 before the token or transitively before the token as it was expanded
541 through a macro. The definition of this flag is very closely defined by
542 the stringizing requirements of the preprocessor.</li>
543 <li><b>DisableExpand</b> - This flag is used internally to the preprocessor to
544 represent identifier tokens which have macro expansion disabled. This
545 prevents them from being considered as candidates for macro expansion ever
546 in the future.</li>
547 <li><b>NeedsCleaning</b> - This flag is set if the original spelling for the
548 token includes a trigraph or escaped newline. Since this is uncommon,
549 many pieces of code can fast-path on tokens that did not need cleaning.
550 </p>
551 </ol>
552</li>
553</ul>
554
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]555<p>One interesting (and somewhat unusual) aspect of normal tokens is that they
556don't contain any semantic information about the lexed value. For example, if
557the token was a pp-number token, we do not represent the value of the number
558that was lexed (this is left for later pieces of code to decide). Additionally,
559the lexer library has no notion of typedef names vs variable names: both are
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]560returned as identifiers, and the parser is left to decide whether a specific
561identifier is a typedef or a variable (tracking this requires scope information
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]562among other things). The parser can do this translation by replacing tokens
563returned by the preprocessor with "Annotation Tokens".</p>
564
565<!-- ======================================================================= -->
566<h3 id="AnnotationToken">Annotation Tokens</h3>
567<!-- ======================================================================= -->
568
569<p>Annotation Tokens are tokens that are synthesized by the parser and injected
570into the preprocessor's token stream (replacing existing tokens) to record
571semantic information found by the parser. For example, if "foo" is found to be
572a typedef, the "foo" <tt>tok::identifier</tt> token is replaced with an
573<tt>tok::annot_typename</tt>. This is useful for a couple of reasons: 1) this
574makes it easy to handle qualified type names (e.g. "foo::bar::baz&lt;42&gt;::t")
575in C++ as a single "token" in the parser. 2) if the parser backtracks, the
576reparse does not need to redo semantic analysis to determine whether a token
577sequence is a variable, type, template, etc.</p>
578
579<p>Annotation Tokens are created by the parser and reinjected into the parser's
580token stream (when backtracking is enabled). Because they can only exist in
581tokens that the preprocessor-proper is done with, it doesn't need to keep around
582flags like "start of line" that the preprocessor uses to do its job.
583Additionally, an annotation token may "cover" a sequence of preprocessor tokens
584(e.g. <tt>a::b::c</tt> is five preprocessor tokens). As such, the valid fields
585of an annotation token are different than the fields for a normal token (but
586they are multiplexed into the normal Token fields):</p>
587
588<ul>
589<li><b>SourceLocation "Location"</b> - The SourceLocation for the annotation
590token indicates the first token replaced by the annotation token. In the example
591above, it would be the location of the "a" identifier.</li>
592
593<li><b>SourceLocation "AnnotationEndLoc"</b> - This holds the location of the
594last token replaced with the annotation token. In the example above, it would
595be the location of the "c" identifier.</li>
596
597<li><b>void* "AnnotationValue"</b> - This contains an opaque object that the
598parser gets from Sema through an Actions module, it is passed around and Sema
599intepretes it, based on the type of annotation token.</li>
600
601<li><b>TokenKind "Kind"</b> - This indicates the kind of Annotation token this
602is. See below for the different valid kinds.</li>
603</ul>
604
605<p>Annotation tokens currently come in three kinds:</p>
606
607<ol>
608<li><b>tok::annot_typename</b>: This annotation token represents a
609resolved typename token that is potentially qualified. The AnnotationValue
Steve Naroffb43a50f2009-01-28 19:39:02 +0000[diff] [blame]610field contains a pointer returned by Action::getTypeName(). In the case of the
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]611Sema actions module, this is a <tt>Decl*</tt> for the type.</li>
612
613<li><b>tok::annot_cxxscope</b>: This annotation token represents a C++ scope
614specifier, such as "A::B::". This corresponds to the grammar productions "::"
615and ":: [opt] nested-name-specifier". The AnnotationValue pointer is returned
616by the Action::ActOnCXXGlobalScopeSpecifier and
617Action::ActOnCXXNestedNameSpecifier callbacks. In the case of Sema, this is a
618<tt>DeclContext*</tt>.</li>
619
620<li><b>tok::annot_template_id</b>: This annotation token represents a C++
621template-id such as "foo&lt;int, 4&gt;", which may refer to a function or type
622depending on whether foo is a function template or class template. The
623AnnotationValue pointer is a pointer to a malloc'd TemplateIdAnnotation object.
624FIXME: I don't think the parsing logic is right for this. Shouldn't type
625templates be turned into annot_typename??</li>
626
627</ol>
628
Cedric Venetda76b282009-01-06 16:22:54 +0000[diff] [blame]629<p>As mentioned above, annotation tokens are not returned by the preprocessor,
Chris Lattner3932fe02009-01-06 06:02:08 +0000[diff] [blame]630they are formed on demand by the parser. This means that the parser has to be
631aware of cases where an annotation could occur and form it where appropriate.
632This is somewhat similar to how the parser handles Translation Phase 6 of C99:
633String Concatenation (see C99 5.1.1.2). In the case of string concatenation,
634the preprocessor just returns distinct tok::string_literal and
635tok::wide_string_literal tokens and the parser eats a sequence of them wherever
636the grammar indicates that a string literal can occur.</p>
637
638<p>In order to do this, whenever the parser expects a tok::identifier or
639tok::coloncolon, it should call the TryAnnotateTypeOrScopeToken or
640TryAnnotateCXXScopeToken methods to form the annotation token. These methods
641will maximally form the specified annotation tokens and replace the current
642token with them, if applicable. If the current tokens is not valid for an
643annotation token, it will remain an identifier or :: token.</p>
644
645
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]646
647<!-- ======================================================================= -->
648<h3 id="Lexer">The Lexer class</h3>
649<!-- ======================================================================= -->
650
651<p>The Lexer class provides the mechanics of lexing tokens out of a source
652buffer and deciding what they mean. The Lexer is complicated by the fact that
653it operates on raw buffers that have not had spelling eliminated (this is a
654necessity to get decent performance), but this is countered with careful coding
655as well as standard performance techniques (for example, the comment handling
656code is vectorized on X86 and PowerPC hosts).</p>
657
658<p>The lexer has a couple of interesting modal features:</p>
659
660<ul>
661<li>The lexer can operate in 'raw' mode. This mode has several features that
662 make it possible to quickly lex the file (e.g. it stops identifier lookup,
663 doesn't specially handle preprocessor tokens, handles EOF differently, etc).
664 This mode is used for lexing within an "<tt>#if 0</tt>" block, for
665 example.</li>
666<li>The lexer can capture and return comments as tokens. This is required to
667 support the -C preprocessor mode, which passes comments through, and is
668 used by the diagnostic checker to identifier expect-error annotations.</li>
669<li>The lexer can be in ParsingFilename mode, which happens when preprocessing
Chris Lattner84386242007-09-16 19:25:23 +0000[diff] [blame]670 after reading a #include directive. This mode changes the parsing of '&lt;'
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]671 to return an "angled string" instead of a bunch of tokens for each thing
672 within the filename.</li>
673<li>When parsing a preprocessor directive (after "<tt>#</tt>") the
674 ParsingPreprocessorDirective mode is entered. This changes the parser to
675 return EOM at a newline.</li>
676<li>The Lexer uses a LangOptions object to know whether trigraphs are enabled,
677 whether C++ or ObjC keywords are recognized, etc.</li>
678</ul>
679
680<p>In addition to these modes, the lexer keeps track of a couple of other
681 features that are local to a lexed buffer, which change as the buffer is
682 lexed:</p>
683
684<ul>
685<li>The Lexer uses BufferPtr to keep track of the current character being
686 lexed.</li>
687<li>The Lexer uses IsAtStartOfLine to keep track of whether the next lexed token
688 will start with its "start of line" bit set.</li>
689<li>The Lexer keeps track of the current #if directives that are active (which
690 can be nested).</li>
691<li>The Lexer keeps track of an <a href="#MultipleIncludeOpt">
692 MultipleIncludeOpt</a> object, which is used to
693 detect whether the buffer uses the standard "<tt>#ifndef XX</tt> /
694 <tt>#define XX</tt>" idiom to prevent multiple inclusion. If a buffer does,
695 subsequent includes can be ignored if the XX macro is defined.</li>
696</ul>
697
698<!-- ======================================================================= -->
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]699<h3 id="TokenLexer">The TokenLexer class</h3>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]700<!-- ======================================================================= -->
701
Chris Lattner79281252008-03-09 02:27:26 +0000[diff] [blame]702<p>The TokenLexer class is a token provider that returns tokens from a list
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]703of tokens that came from somewhere else. It typically used for two things: 1)
704returning tokens from a macro definition as it is being expanded 2) returning
705tokens from an arbitrary buffer of tokens. The later use is used by _Pragma and
706will most likely be used to handle unbounded look-ahead for the C++ parser.</p>
707
708<!-- ======================================================================= -->
709<h3 id="MultipleIncludeOpt">The MultipleIncludeOpt class</h3>
710<!-- ======================================================================= -->
711
712<p>The MultipleIncludeOpt class implements a really simple little state machine
713that is used to detect the standard "<tt>#ifndef XX</tt> / <tt>#define XX</tt>"
714idiom that people typically use to prevent multiple inclusion of headers. If a
715buffer uses this idiom and is subsequently #include'd, the preprocessor can
716simply check to see whether the guarding condition is defined or not. If so,
717the preprocessor can completely ignore the include of the header.</p>
718
719
720
721<!-- ======================================================================= -->
722<h2 id="libparse">The Parser Library</h2>
723<!-- ======================================================================= -->
724
725<!-- ======================================================================= -->
726<h2 id="libast">The AST Library</h2>
727<!-- ======================================================================= -->
728
729<!-- ======================================================================= -->
730<h3 id="Type">The Type class and its subclasses</h3>
731<!-- ======================================================================= -->
732
733<p>The Type class (and its subclasses) are an important part of the AST. Types
734are accessed through the ASTContext class, which implicitly creates and uniques
735them as they are needed. Types have a couple of non-obvious features: 1) they
736do not capture type qualifiers like const or volatile (See
737<a href="#QualType">QualType</a>), and 2) they implicitly capture typedef
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]738information. Once created, types are immutable (unlike decls).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]739
740<p>Typedefs in C make semantic analysis a bit more complex than it would
741be without them. The issue is that we want to capture typedef information
742and represent it in the AST perfectly, but the semantics of operations need to
743"see through" typedefs. For example, consider this code:</p>
744
745<code>
746void func() {<br>
Bill Wendling30d17752007-10-06 01:56:01 +0000[diff] [blame]747&nbsp;&nbsp;typedef int foo;<br>
748&nbsp;&nbsp;foo X, *Y;<br>
749&nbsp;&nbsp;typedef foo* bar;<br>
750&nbsp;&nbsp;bar Z;<br>
751&nbsp;&nbsp;*X; <i>// error</i><br>
752&nbsp;&nbsp;**Y; <i>// error</i><br>
753&nbsp;&nbsp;**Z; <i>// error</i><br>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]754}<br>
755</code>
756
757<p>The code above is illegal, and thus we expect there to be diagnostics emitted
758on the annotated lines. In this example, we expect to get:</p>
759
760<pre>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]761<b>test.c:6:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]762*X; // error
763<font color="blue">^~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]764<b>test.c:7:1: error: indirection requires pointer operand ('foo' invalid)</b>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]765**Y; // error
766<font color="blue">^~~</font>
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]767<b>test.c:8:1: error: indirection requires pointer operand ('foo' invalid)</b>
768**Z; // error
769<font color="blue">^~~</font>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]770</pre>
771
772<p>While this example is somewhat silly, it illustrates the point: we want to
773retain typedef information where possible, so that we can emit errors about
774"<tt>std::string</tt>" instead of "<tt>std::basic_string&lt;char, std:...</tt>".
775Doing this requires properly keeping typedef information (for example, the type
776of "X" is "foo", not "int"), and requires properly propagating it through the
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]777various operators (for example, the type of *Y is "foo", not "int"). In order
778to retain this information, the type of these expressions is an instance of the
779TypedefType class, which indicates that the type of these expressions is a
780typedef for foo.
781</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]782
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]783<p>Representing types like this is great for diagnostics, because the
784user-specified type is always immediately available. There are two problems
785with this: first, various semantic checks need to make judgements about the
Chris Lattner33fc68a2007-07-31 18:54:50 +0000[diff] [blame]786<em>actual structure</em> of a type, ignoring typdefs. Second, we need an
787efficient way to query whether two types are structurally identical to each
788other, ignoring typedefs. The solution to both of these problems is the idea of
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]789canonical types.</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]790
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]791<!-- =============== -->
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]792<h4>Canonical Types</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]793<!-- =============== -->
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]794
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]795<p>Every instance of the Type class contains a canonical type pointer. For
796simple types with no typedefs involved (e.g. "<tt>int</tt>", "<tt>int*</tt>",
797"<tt>int**</tt>"), the type just points to itself. For types that have a
798typedef somewhere in their structure (e.g. "<tt>foo</tt>", "<tt>foo*</tt>",
799"<tt>foo**</tt>", "<tt>bar</tt>"), the canonical type pointer points to their
800structurally equivalent type without any typedefs (e.g. "<tt>int</tt>",
801"<tt>int*</tt>", "<tt>int**</tt>", and "<tt>int*</tt>" respectively).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]802
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]803<p>This design provides a constant time operation (dereferencing the canonical
804type pointer) that gives us access to the structure of types. For example,
805we can trivially tell that "bar" and "foo*" are the same type by dereferencing
806their canonical type pointers and doing a pointer comparison (they both point
807to the single "<tt>int*</tt>" type).</p>
808
809<p>Canonical types and typedef types bring up some complexities that must be
810carefully managed. Specifically, the "isa/cast/dyncast" operators generally
811shouldn't be used in code that is inspecting the AST. For example, when type
812checking the indirection operator (unary '*' on a pointer), the type checker
813must verify that the operand has a pointer type. It would not be correct to
814check that with "<tt>isa&lt;PointerType&gt;(SubExpr-&gt;getType())</tt>",
815because this predicate would fail if the subexpression had a typedef type.</p>
816
817<p>The solution to this problem are a set of helper methods on Type, used to
818check their properties. In this case, it would be correct to use
819"<tt>SubExpr-&gt;getType()-&gt;isPointerType()</tt>" to do the check. This
820predicate will return true if the <em>canonical type is a pointer</em>, which is
821true any time the type is structurally a pointer type. The only hard part here
822is remembering not to use the <tt>isa/cast/dyncast</tt> operations.</p>
823
824<p>The second problem we face is how to get access to the pointer type once we
825know it exists. To continue the example, the result type of the indirection
826operator is the pointee type of the subexpression. In order to determine the
827type, we need to get the instance of PointerType that best captures the typedef
828information in the program. If the type of the expression is literally a
829PointerType, we can return that, otherwise we have to dig through the
830typedefs to find the pointer type. For example, if the subexpression had type
831"<tt>foo*</tt>", we could return that type as the result. If the subexpression
832had type "<tt>bar</tt>", we want to return "<tt>foo*</tt>" (note that we do
833<em>not</em> want "<tt>int*</tt>"). In order to provide all of this, Type has
Chris Lattner11406c12007-07-31 16:50:51 +0000[diff] [blame]834a getAsPointerType() method that checks whether the type is structurally a
Chris Lattner8a2bc622007-07-31 06:37:39 +0000[diff] [blame]835PointerType and, if so, returns the best one. If not, it returns a null
836pointer.</p>
837
838<p>This structure is somewhat mystical, but after meditating on it, it will
839make sense to you :).</p>
Chris Lattner86920d32007-07-31 05:42:17 +0000[diff] [blame]840
841<!-- ======================================================================= -->
842<h3 id="QualType">The QualType class</h3>
843<!-- ======================================================================= -->
844
845<p>The QualType class is designed as a trivial value class that is small,
846passed by-value and is efficient to query. The idea of QualType is that it
847stores the type qualifiers (const, volatile, restrict) separately from the types
848themselves: QualType is conceptually a pair of "Type*" and bits for the type
849qualifiers.</p>
850
851<p>By storing the type qualifiers as bits in the conceptual pair, it is
852extremely efficient to get the set of qualifiers on a QualType (just return the
853field of the pair), add a type qualifier (which is a trivial constant-time
854operation that sets a bit), and remove one or more type qualifiers (just return
855a QualType with the bitfield set to empty).</p>
856
857<p>Further, because the bits are stored outside of the type itself, we do not
858need to create duplicates of types with different sets of qualifiers (i.e. there
859is only a single heap allocated "int" type: "const int" and "volatile const int"
860both point to the same heap allocated "int" type). This reduces the heap size
861used to represent bits and also means we do not have to consider qualifiers when
862uniquing types (<a href="#Type">Type</a> does not even contain qualifiers).</p>
863
864<p>In practice, on hosts where it is safe, the 3 type qualifiers are stored in
865the low bit of the pointer to the Type object. This means that QualType is
866exactly the same size as a pointer, and this works fine on any system where
867malloc'd objects are at least 8 byte aligned.</p>
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]868
869<!-- ======================================================================= -->
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]870<h3 id="DeclarationName">Declaration names</h3>
871<!-- ======================================================================= -->
872
873<p>The <tt>DeclarationName</tt> class represents the name of a
874 declaration in Clang. Declarations in the C family of languages can
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]875 take several different forms. Most declarations are named by
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]876 simple identifiers, e.g., "<code>f</code>" and "<code>x</code>" in
877 the function declaration <code>f(int x)</code>. In C++, declaration
878 names can also name class constructors ("<code>Class</code>"
879 in <code>struct Class { Class(); }</code>), class destructors
880 ("<code>~Class</code>"), overloaded operator names ("operator+"),
881 and conversion functions ("<code>operator void const *</code>"). In
882 Objective-C, declaration names can refer to the names of Objective-C
883 methods, which involve the method name and the parameters,
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]884 collectively called a <i>selector</i>, e.g.,
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]885 "<code>setWidth:height:</code>". Since all of these kinds of
Chris Lattner3fcbb892008-11-23 08:32:53 +0000[diff] [blame]886 entities - variables, functions, Objective-C methods, C++
887 constructors, destructors, and operators - are represented as
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]888 subclasses of Clang's common <code>NamedDecl</code>
889 class, <code>DeclarationName</code> is designed to efficiently
890 represent any kind of name.</p>
891
892<p>Given
893 a <code>DeclarationName</code> <code>N</code>, <code>N.getNameKind()</code>
Douglas Gregor2def4832008-11-17 20:34:05 +0000[diff] [blame]894 will produce a value that describes what kind of name <code>N</code>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]895 stores. There are 8 options (all of the names are inside
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]896 the <code>DeclarationName</code> class)</p>
897<dl>
898 <dt>Identifier</dt>
899 <dd>The name is a simple
900 identifier. Use <code>N.getAsIdentifierInfo()</code> to retrieve the
901 corresponding <code>IdentifierInfo*</code> pointing to the actual
902 identifier. Note that C++ overloaded operators (e.g.,
903 "<code>operator+</code>") are represented as special kinds of
904 identifiers. Use <code>IdentifierInfo</code>'s <code>getOverloadedOperatorID</code>
905 function to determine whether an identifier is an overloaded
906 operator name.</dd>
907
908 <dt>ObjCZeroArgSelector, ObjCOneArgSelector,
909 ObjCMultiArgSelector</dt>
910 <dd>The name is an Objective-C selector, which can be retrieved as a
911 <code>Selector</code> instance
912 via <code>N.getObjCSelector()</code>. The three possible name
913 kinds for Objective-C reflect an optimization within
914 the <code>DeclarationName</code> class: both zero- and
915 one-argument selectors are stored as a
916 masked <code>IdentifierInfo</code> pointer, and therefore require
917 very little space, since zero- and one-argument selectors are far
918 more common than multi-argument selectors (which use a different
919 structure).</dd>
920
921 <dt>CXXConstructorName</dt>
922 <dd>The name is a C++ constructor
923 name. Use <code>N.getCXXNameType()</code> to retrieve
924 the <a href="#QualType">type</a> that this constructor is meant to
925 construct. The type is always the canonical type, since all
926 constructors for a given type have the same name.</dd>
927
928 <dt>CXXDestructorName</dt>
929 <dd>The name is a C++ destructor
930 name. Use <code>N.getCXXNameType()</code> to retrieve
931 the <a href="#QualType">type</a> whose destructor is being
932 named. This type is always a canonical type.</dd>
933
934 <dt>CXXConversionFunctionName</dt>
935 <dd>The name is a C++ conversion function. Conversion functions are
936 named according to the type they convert to, e.g., "<code>operator void
937 const *</code>". Use <code>N.getCXXNameType()</code> to retrieve
938 the type that this conversion function converts to. This type is
939 always a canonical type.</dd>
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]940
941 <dt>CXXOperatorName</dt>
942 <dd>The name is a C++ overloaded operator name. Overloaded operators
943 are named according to their spelling, e.g.,
944 "<code>operator+</code>" or "<code>operator new
945 []</code>". Use <code>N.getCXXOverloadedOperator()</code> to
946 retrieve the overloaded operator (a value of
947 type <code>OverloadedOperatorKind</code>).</dd>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]948</dl>
949
950<p><code>DeclarationName</code>s are cheap to create, copy, and
951 compare. They require only a single pointer's worth of storage in
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]952 the common cases (identifiers, zero-
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]953 and one-argument Objective-C selectors) and use dense, uniqued
954 storage for the other kinds of
955 names. Two <code>DeclarationName</code>s can be compared for
956 equality (<code>==</code>, <code>!=</code>) using a simple bitwise
957 comparison, can be ordered
958 with <code>&lt;</code>, <code>&gt;</code>, <code>&lt;=</code>,
959 and <code>&gt;=</code> (which provide a lexicographical ordering for
960 normal identifiers but an unspecified ordering for other kinds of
961 names), and can be placed into LLVM <code>DenseMap</code>s
962 and <code>DenseSet</code>s.</p>
963
964<p><code>DeclarationName</code> instances can be created in different
965 ways depending on what kind of name the instance will store. Normal
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]966 identifiers (<code>IdentifierInfo</code> pointers) and Objective-C selectors
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]967 (<code>Selector</code>) can be implicitly converted
968 to <code>DeclarationName</code>s. Names for C++ constructors,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]969 destructors, conversion functions, and overloaded operators can be retrieved from
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]970 the <code>DeclarationNameTable</code>, an instance of which is
971 available as <code>ASTContext::DeclarationNames</code>. The member
972 functions <code>getCXXConstructorName</code>, <code>getCXXDestructorName</code>,
Douglas Gregore94ca9e42008-11-18 14:39:36 +0000[diff] [blame]973 <code>getCXXConversionFunctionName</code>, and <code>getCXXOperatorName</code>, respectively,
974 return <code>DeclarationName</code> instances for the four kinds of
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]975 C++ special function names.</p>
976
977<!-- ======================================================================= -->
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]978<h3 id="DeclContext">Declaration contexts</h3>
979<!-- ======================================================================= -->
980<p>Every declaration in a program exists within some <i>declaration
981 context</i>, such as a translation unit, namespace, class, or
982 function. Declaration contexts in Clang are represented by
983 the <code>DeclContext</code> class, from which the various
984 declaration-context AST nodes
985 (<code>TranslationUnitDecl</code>, <code>NamespaceDecl</code>, <code>RecordDecl</code>, <code>FunctionDecl</code>,
986 etc.) will derive. The <code>DeclContext</code> class provides
987 several facilities common to each declaration context:</p>
988<dl>
989 <dt>Source-centric vs. Semantics-centric View of Declarations</dt>
990 <dd><code>DeclContext</code> provides two views of the declarations
991 stored within a declaration context. The source-centric view
992 accurately represents the program source code as written, including
993 multiple declarations of entities where present (see the
994 section <a href="#Redeclarations">Redeclarations and
995 Overloads</a>), while the semantics-centric view represents the
996 program semantics. The two views are kept synchronized by semantic
997 analysis while the ASTs are being constructed.</dd>
998
999 <dt>Storage of declarations within that context</dt>
1000 <dd>Every declaration context can contain some number of
1001 declarations. For example, a C++ class (represented
1002 by <code>RecordDecl</code>) contains various member functions,
1003 fields, nested types, and so on. All of these declarations will be
1004 stored within the <code>DeclContext</code>, and one can iterate
1005 over the declarations via
1006 [<code>DeclContext::decls_begin()</code>,
1007 <code>DeclContext::decls_end()</code>). This mechanism provides
1008 the source-centric view of declarations in the context.</dd>
1009
1010 <dt>Lookup of declarations within that context</dt>
1011 <dd>The <code>DeclContext</code> structure provides efficient name
1012 lookup for names within that declaration context. For example,
1013 if <code>N</code> is a namespace we can look for the
1014 name <code>N::f</code>
1015 using <code>DeclContext::lookup</code>. The lookup itself is
1016 based on a lazily-constructed array (for declaration contexts
1017 with a small number of declarations) or hash table (for
1018 declaration contexts with more declarations). The lookup
1019 operation provides the semantics-centric view of the declarations
1020 in the context.</dd>
1021
1022 <dt>Ownership of declarations</dt>
1023 <dd>The <code>DeclContext</code> owns all of the declarations that
1024 were declared within its declaration context, and is responsible
1025 for the management of their memory as well as their
1026 (de-)serialization.</dd>
1027</dl>
1028
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1029<p>All declarations are stored within a declaration context, and one
1030 can query
1031 information about the context in which each declaration lives. One
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1032 can retrieve the <code>DeclContext</code> that contains a
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1033 particular <code>Decl</code>
1034 using <code>Decl::getDeclContext</code>. However, see the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1035 section <a href="#LexicalAndSemanticContexts">Lexical and Semantic
1036 Contexts</a> for more information about how to interpret this
1037 context information.</p>
1038
1039<h4 id="Redeclarations">Redeclarations and Overloads</h4>
1040<p>Within a translation unit, it is common for an entity to be
1041declared several times. For example, we might declare a function "f"
1042 and then later re-declare it as part of an inlined definition:</p>
1043
1044<pre>
1045void f(int x, int y, int z = 1);
1046
1047inline void f(int x, int y, int z) { /* ... */ }
1048</pre>
1049
1050<p>The representation of "f" differs in the source-centric and
1051 semantics-centric views of a declaration context. In the
1052 source-centric view, all redeclarations will be present, in the
1053 order they occurred in the source code, making
1054 this view suitable for clients that wish to see the structure of
1055 the source code. In the semantics-centric view, only the most recent "f"
1056 will be found by the lookup, since it effectively replaces the first
1057 declaration of "f".</p>
1058
1059<p>In the semantics-centric view, overloading of functions is
1060 represented explicitly. For example, given two declarations of a
1061 function "g" that are overloaded, e.g.,</p>
1062<pre>
1063void g();
1064void g(int);
1065</pre>
1066<p>the <code>DeclContext::lookup</code> operation will return
1067 an <code>OverloadedFunctionDecl</code> that contains both
1068 declarations of "g". Clients that perform semantic analysis on a
1069 program that is not concerned with the actual source code will
1070 primarily use this semantics-centric view.</p>
1071
1072<h4 id="LexicalAndSemanticContexts">Lexical and Semantic Contexts</h4>
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1073<p>Each declaration has two potentially different
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1074 declaration contexts: a <i>lexical</i> context, which corresponds to
1075 the source-centric view of the declaration context, and
1076 a <i>semantic</i> context, which corresponds to the
1077 semantics-centric view. The lexical context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1078 via <code>Decl::getLexicalDeclContext</code> while the
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1079 semantic context is accessible
Douglas Gregor4afa39d2009-01-20 01:17:11 +0000[diff] [blame]1080 via <code>Decl::getDeclContext</code>, both of which return
Douglas Gregor074149e2009-01-05 19:45:36 +0000[diff] [blame]1081 <code>DeclContext</code> pointers. For most declarations, the two
1082 contexts are identical. For example:</p>
1083
1084<pre>
1085class X {
1086public:
1087 void f(int x);
1088};
1089</pre>
1090
1091<p>Here, the semantic and lexical contexts of <code>X::f</code> are
1092 the <code>DeclContext</code> associated with the
1093 class <code>X</code> (itself stored as a <code>RecordDecl</code> AST
1094 node). However, we can now define <code>X::f</code> out-of-line:</p>
1095
1096<pre>
1097void X::f(int x = 17) { /* ... */ }
1098</pre>
1099
1100<p>This definition of has different lexical and semantic
1101 contexts. The lexical context corresponds to the declaration
1102 context in which the actual declaration occurred in the source
1103 code, e.g., the translation unit containing <code>X</code>. Thus,
1104 this declaration of <code>X::f</code> can be found by traversing
1105 the declarations provided by
1106 [<code>decls_begin()</code>, <code>decls_end()</code>) in the
1107 translation unit.</p>
1108
1109<p>The semantic context of <code>X::f</code> corresponds to the
1110 class <code>X</code>, since this member function is (semantically) a
1111 member of <code>X</code>. Lookup of the name <code>f</code> into
1112 the <code>DeclContext</code> associated with <code>X</code> will
1113 then return the definition of <code>X::f</code> (including
1114 information about the default argument).</p>
1115
1116<h4 id="TransparentContexts">Transparent Declaration Contexts</h4>
1117<p>In C and C++, there are several contexts in which names that are
1118 logically declared inside another declaration will actually "leak"
1119 out into the enclosing scope from the perspective of name
1120 lookup. The most obvious instance of this behavior is in
1121 enumeration types, e.g.,</p>
1122<pre>
1123enum Color {
1124 Red,
1125 Green,
1126 Blue
1127};
1128</pre>
1129
1130<p>Here, <code>Color</code> is an enumeration, which is a declaration
1131 context that contains the
1132 enumerators <code>Red</code>, <code>Green</code>,
1133 and <code>Blue</code>. Thus, traversing the list of declarations
1134 contained in the enumeration <code>Color</code> will
1135 yield <code>Red</code>, <code>Green</code>,
1136 and <code>Blue</code>. However, outside of the scope
1137 of <code>Color</code> one can name the enumerator <code>Red</code>
1138 without qualifying the name, e.g.,</p>
1139
1140<pre>
1141Color c = Red;
1142</pre>
1143
1144<p>There are other entities in C++ that provide similar behavior. For
1145 example, linkage specifications that use curly braces:</p>
1146
1147<pre>
1148extern "C" {
1149 void f(int);
1150 void g(int);
1151}
1152// f and g are visible here
1153</pre>
1154
1155<p>For source-level accuracy, we treat the linkage specification and
1156 enumeration type as a
1157 declaration context in which its enclosed declarations ("Red",
1158 "Green", and "Blue"; "f" and "g")
1159 are declared. However, these declarations are visible outside of the
1160 scope of the declaration context.</p>
1161
1162<p>These language features (and several others, described below) have
1163 roughly the same set of
1164 requirements: declarations are declared within a particular lexical
1165 context, but the declarations are also found via name lookup in
1166 scopes enclosing the declaration itself. This feature is implemented
1167 via <i>transparent</i> declaration contexts
1168 (see <code>DeclContext::isTransparentContext()</code>), whose
1169 declarations are visible in the nearest enclosing non-transparent
1170 declaration context. This means that the lexical context of the
1171 declaration (e.g., an enumerator) will be the
1172 transparent <code>DeclContext</code> itself, as will the semantic
1173 context, but the declaration will be visible in every outer context
1174 up to and including the first non-transparent declaration context (since
1175 transparent declaration contexts can be nested).</p>
1176
1177<p>The transparent <code>DeclContexts</code> are:</p>
1178<ul>
1179 <li>Enumerations (but not C++0x "scoped enumerations"):
1180 <pre>
1181enum Color {
1182 Red,
1183 Green,
1184 Blue
1185};
1186// Red, Green, and Blue are in scope
1187 </pre></li>
1188 <li>C++ linkage specifications:
1189 <pre>
1190extern "C" {
1191 void f(int);
1192 void g(int);
1193}
1194// f and g are in scope
1195 </pre></li>
1196 <li>Anonymous unions and structs:
1197 <pre>
1198struct LookupTable {
1199 bool IsVector;
1200 union {
1201 std::vector&lt;Item&gt; *Vector;
1202 std::set&lt;Item&gt; *Set;
1203 };
1204};
1205
1206LookupTable LT;
1207LT.Vector = 0; // Okay: finds Vector inside the unnamed union
1208 </pre>
1209 </li>
1210 <li>C++0x inline namespaces:
1211<pre>
1212namespace mylib {
1213 inline namespace debug {
1214 class X;
1215 }
1216}
1217mylib::X *xp; // okay: mylib::X refers to mylib::debug::X
1218</pre>
1219</li>
1220</ul>
1221
1222
1223<h4 id="MultiDeclContext">Multiply-Defined Declaration Contexts</h4>
1224<p>C++ namespaces have the interesting--and, so far, unique--property that
1225the namespace can be defined multiple times, and the declarations
1226provided by each namespace definition are effectively merged (from
1227the semantic point of view). For example, the following two code
1228snippets are semantically indistinguishable:</p>
1229<pre>
1230// Snippet #1:
1231namespace N {
1232 void f();
1233}
1234namespace N {
1235 void f(int);
1236}
1237
1238// Snippet #2:
1239namespace N {
1240 void f();
1241 void f(int);
1242}
1243</pre>
1244
1245<p>In Clang's representation, the source-centric view of declaration
1246 contexts will actually have two separate <code>NamespaceDecl</code>
1247 nodes in Snippet #1, each of which is a declaration context that
1248 contains a single declaration of "f". However, the semantics-centric
1249 view provided by name lookup into the namespace <code>N</code> for
1250 "f" will return an <code>OverloadedFunctionDecl</code> that contains
1251 both declarations of "f".</p>
1252
1253<p><code>DeclContext</code> manages multiply-defined declaration
1254 contexts internally. The
1255 function <code>DeclContext::getPrimaryContext</code> retrieves the
1256 "primary" context for a given <code>DeclContext</code> instance,
1257 which is the <code>DeclContext</code> responsible for maintaining
1258 the lookup table used for the semantics-centric view. Given the
1259 primary context, one can follow the chain
1260 of <code>DeclContext</code> nodes that define additional
1261 declarations via <code>DeclContext::getNextContext</code>. Note that
1262 these functions are used internally within the lookup and insertion
1263 methods of the <code>DeclContext</code>, so the vast majority of
1264 clients can ignore them.</p>
1265
1266<!-- ======================================================================= -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1267<h3 id="CFG">The <tt>CFG</tt> class</h3>
1268<!-- ======================================================================= -->
1269
1270<p>The <tt>CFG</tt> class is designed to represent a source-level
1271control-flow graph for a single statement (<tt>Stmt*</tt>). Typically
1272instances of <tt>CFG</tt> are constructed for function bodies (usually
1273an instance of <tt>CompoundStmt</tt>), but can also be instantiated to
1274represent the control-flow of any class that subclasses <tt>Stmt</tt>,
1275which includes simple expressions. Control-flow graphs are especially
1276useful for performing
1277<a href="http://en.wikipedia.org/wiki/Data_flow_analysis#Sensitivities">flow-
1278or path-sensitive</a> program analyses on a given function.</p>
1279
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1280<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1281<h4>Basic Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1282<!-- ============ -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1283
1284<p>Concretely, an instance of <tt>CFG</tt> is a collection of basic
1285blocks. Each basic block is an instance of <tt>CFGBlock</tt>, which
1286simply contains an ordered sequence of <tt>Stmt*</tt> (each referring
1287to statements in the AST). The ordering of statements within a block
1288indicates unconditional flow of control from one statement to the
1289next. <a href="#ConditionalControlFlow">Conditional control-flow</a>
1290is represented using edges between basic blocks. The statements
1291within a given <tt>CFGBlock</tt> can be traversed using
1292the <tt>CFGBlock::*iterator</tt> interface.</p>
1293
1294<p>
Ted Kremenek18e17e72007-10-18 22:50:52 +0000[diff] [blame]1295A <tt>CFG</tt> object owns the instances of <tt>CFGBlock</tt> within
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1296the control-flow graph it represents. Each <tt>CFGBlock</tt> within a
1297CFG is also uniquely numbered (accessible
1298via <tt>CFGBlock::getBlockID()</tt>). Currently the number is
1299based on the ordering the blocks were created, but no assumptions
1300should be made on how <tt>CFGBlock</tt>s are numbered other than their
1301numbers are unique and that they are numbered from 0..N-1 (where N is
1302the number of basic blocks in the CFG).</p>
1303
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1304<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1305<h4>Entry and Exit Blocks</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1306<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1307
1308Each instance of <tt>CFG</tt> contains two special blocks:
1309an <i>entry</i> block (accessible via <tt>CFG::getEntry()</tt>), which
1310has no incoming edges, and an <i>exit</i> block (accessible
1311via <tt>CFG::getExit()</tt>), which has no outgoing edges. Neither
1312block contains any statements, and they serve the role of providing a
1313clear entrance and exit for a body of code such as a function body.
1314The presence of these empty blocks greatly simplifies the
1315implementation of many analyses built on top of CFGs.
1316
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1317<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1318<h4 id ="ConditionalControlFlow">Conditional Control-Flow</h4>
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1319<!-- ===================================================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1320
1321<p>Conditional control-flow (such as those induced by if-statements
1322and loops) is represented as edges between <tt>CFGBlock</tt>s.
1323Because different C language constructs can induce control-flow,
1324each <tt>CFGBlock</tt> also records an extra <tt>Stmt*</tt> that
1325represents the <i>terminator</i> of the block. A terminator is simply
1326the statement that caused the control-flow, and is used to identify
1327the nature of the conditional control-flow between blocks. For
1328example, in the case of an if-statement, the terminator refers to
1329the <tt>IfStmt</tt> object in the AST that represented the given
1330branch.</p>
1331
1332<p>To illustrate, consider the following code example:</p>
1333
1334<code>
1335int foo(int x) {<br>
1336&nbsp;&nbsp;x = x + 1;<br>
1337<br>
1338&nbsp;&nbsp;if (x > 2) x++;<br>
1339&nbsp;&nbsp;else {<br>
1340&nbsp;&nbsp;&nbsp;&nbsp;x += 2;<br>
1341&nbsp;&nbsp;&nbsp;&nbsp;x *= 2;<br>
1342&nbsp;&nbsp;}<br>
1343<br>
1344&nbsp;&nbsp;return x;<br>
1345}
1346</code>
1347
1348<p>After invoking the parser+semantic analyzer on this code fragment,
1349the AST of the body of <tt>foo</tt> is referenced by a
1350single <tt>Stmt*</tt>. We can then construct an instance
1351of <tt>CFG</tt> representing the control-flow graph of this function
1352body by single call to a static class method:</p>
1353
1354<code>
1355&nbsp;&nbsp;Stmt* FooBody = ...<br>
1356&nbsp;&nbsp;CFG* FooCFG = <b>CFG::buildCFG</b>(FooBody);
1357</code>
1358
1359<p>It is the responsibility of the caller of <tt>CFG::buildCFG</tt>
1360to <tt>delete</tt> the returned <tt>CFG*</tt> when the CFG is no
1361longer needed.</p>
1362
1363<p>Along with providing an interface to iterate over
1364its <tt>CFGBlock</tt>s, the <tt>CFG</tt> class also provides methods
1365that are useful for debugging and visualizing CFGs. For example, the
1366method
1367<tt>CFG::dump()</tt> dumps a pretty-printed version of the CFG to
1368standard error. This is especially useful when one is using a
1369debugger such as gdb. For example, here is the output
1370of <tt>FooCFG->dump()</tt>:</p>
1371
1372<code>
1373&nbsp;[ B5 (ENTRY) ]<br>
1374&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (0):<br>
1375&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B4<br>
1376<br>
1377&nbsp;[ B4 ]<br>
1378&nbsp;&nbsp;&nbsp;&nbsp;1: x = x + 1<br>
1379&nbsp;&nbsp;&nbsp;&nbsp;2: (x > 2)<br>
1380&nbsp;&nbsp;&nbsp;&nbsp;<b>T: if [B4.2]</b><br>
1381&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B5<br>
1382&nbsp;&nbsp;&nbsp;&nbsp;Successors (2): B3 B2<br>
1383<br>
1384&nbsp;[ B3 ]<br>
1385&nbsp;&nbsp;&nbsp;&nbsp;1: x++<br>
1386&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1387&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1388<br>
1389&nbsp;[ B2 ]<br>
1390&nbsp;&nbsp;&nbsp;&nbsp;1: x += 2<br>
1391&nbsp;&nbsp;&nbsp;&nbsp;2: x *= 2<br>
1392&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B4<br>
1393&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B1<br>
1394<br>
1395&nbsp;[ B1 ]<br>
1396&nbsp;&nbsp;&nbsp;&nbsp;1: return x;<br>
1397&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (2): B2 B3<br>
1398&nbsp;&nbsp;&nbsp;&nbsp;Successors (1): B0<br>
1399<br>
1400&nbsp;[ B0 (EXIT) ]<br>
1401&nbsp;&nbsp;&nbsp;&nbsp;Predecessors (1): B1<br>
1402&nbsp;&nbsp;&nbsp;&nbsp;Successors (0):
1403</code>
1404
1405<p>For each block, the pretty-printed output displays for each block
1406the number of <i>predecessor</i> blocks (blocks that have outgoing
1407control-flow to the given block) and <i>successor</i> blocks (blocks
1408that have control-flow that have incoming control-flow from the given
1409block). We can also clearly see the special entry and exit blocks at
1410the beginning and end of the pretty-printed output. For the entry
1411block (block B5), the number of predecessor blocks is 0, while for the
1412exit block (block B0) the number of successor blocks is 0.</p>
1413
1414<p>The most interesting block here is B4, whose outgoing control-flow
1415represents the branching caused by the sole if-statement
1416in <tt>foo</tt>. Of particular interest is the second statement in
1417the block, <b><tt>(x > 2)</tt></b>, and the terminator, printed
1418as <b><tt>if [B4.2]</tt></b>. The second statement represents the
1419evaluation of the condition of the if-statement, which occurs before
1420the actual branching of control-flow. Within the <tt>CFGBlock</tt>
1421for B4, the <tt>Stmt*</tt> for the second statement refers to the
1422actual expression in the AST for <b><tt>(x > 2)</tt></b>. Thus
1423pointers to subclasses of <tt>Expr</tt> can appear in the list of
1424statements in a block, and not just subclasses of <tt>Stmt</tt> that
1425refer to proper C statements.</p>
1426
1427<p>The terminator of block B4 is a pointer to the <tt>IfStmt</tt>
1428object in the AST. The pretty-printer outputs <b><tt>if
1429[B4.2]</tt></b> because the condition expression of the if-statement
1430has an actual place in the basic block, and thus the terminator is
1431essentially
1432<i>referring</i> to the expression that is the second statement of
1433block B4 (i.e., B4.2). In this manner, conditions for control-flow
1434(which also includes conditions for loops and switch statements) are
1435hoisted into the actual basic block.</p>
1436
Chris Lattner62fd2782008-11-22 21:41:31 +0000[diff] [blame]1437<!-- ===================== -->
1438<!-- <h4>Implicit Control-Flow</h4> -->
1439<!-- ===================== -->
Ted Kremenek8bc05712007-10-10 23:01:43 +0000[diff] [blame]1440
1441<!--
1442<p>A key design principle of the <tt>CFG</tt> class was to not require
1443any transformations to the AST in order to represent control-flow.
1444Thus the <tt>CFG</tt> does not perform any "lowering" of the
1445statements in an AST: loops are not transformed into guarded gotos,
1446short-circuit operations are not converted to a set of if-statements,
1447and so on.</p>
1448-->
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1449
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1450
1451<!-- ======================================================================= -->
1452<h3 id="Constants">Constant Folding in the Clang AST</h3>
1453<!-- ======================================================================= -->
1454
1455<p>There are several places where constants and constant folding matter a lot to
1456the Clang front-end. First, in general, we prefer the AST to retain the source
1457code as close to how the user wrote it as possible. This means that if they
1458wrote "5+4", we want to keep the addition and two constants in the AST, we don't
1459want to fold to "9". This means that constant folding in various ways turns
1460into a tree walk that needs to handle the various cases.</p>
1461
1462<p>However, there are places in both C and C++ that require constants to be
1463folded. For example, the C standard defines what an "integer constant
1464expression" (i-c-e) is with very precise and specific requirements. The
1465language then requires i-c-e's in a lot of places (for example, the size of a
1466bitfield, the value for a case statement, etc). For these, we have to be able
1467to constant fold the constants, to do semantic checks (e.g. verify bitfield size
1468is non-negative and that case statements aren't duplicated). We aim for Clang
1469to be very pedantic about this, diagnosing cases when the code does not use an
1470i-c-e where one is required, but accepting the code unless running with
1471<tt>-pedantic-errors</tt>.</p>
1472
1473<p>Things get a little bit more tricky when it comes to compatibility with
1474real-world source code. Specifically, GCC has historically accepted a huge
1475superset of expressions as i-c-e's, and a lot of real world code depends on this
1476unfortuate accident of history (including, e.g., the glibc system headers). GCC
1477accepts anything its "fold" optimizer is capable of reducing to an integer
1478constant, which means that the definition of what it accepts changes as its
1479optimizer does. One example is that GCC accepts things like "case X-X:" even
1480when X is a variable, because it can fold this to 0.</p>
1481
1482<p>Another issue are how constants interact with the extensions we support, such
1483as __builtin_constant_p, __builtin_inf, __extension__ and many others. C99
1484obviously does not specify the semantics of any of these extensions, and the
1485definition of i-c-e does not include them. However, these extensions are often
1486used in real code, and we have to have a way to reason about them.</p>
1487
1488<p>Finally, this is not just a problem for semantic analysis. The code
1489generator and other clients have to be able to fold constants (e.g. to
1490initialize global variables) and has to handle a superset of what C99 allows.
1491Further, these clients can benefit from extended information. For example, we
1492know that "foo()||1" always evaluates to true, but we can't replace the
1493expression with true because it has side effects.</p>
1494
1495<!-- ======================= -->
1496<h4>Implementation Approach</h4>
1497<!-- ======================= -->
1498
1499<p>After trying several different approaches, we've finally converged on a
1500design (Note, at the time of this writing, not all of this has been implemented,
1501consider this a design goal!). Our basic approach is to define a single
1502recursive method evaluation method (<tt>Expr::Evaluate</tt>), which is
1503implemented in <tt>AST/ExprConstant.cpp</tt>. Given an expression with 'scalar'
1504type (integer, fp, complex, or pointer) this method returns the following
1505information:</p>
1506
1507<ul>
1508<li>Whether the expression is an integer constant expression, a general
1509 constant that was folded but has no side effects, a general constant that
1510 was folded but that does have side effects, or an uncomputable/unfoldable
1511 value.
1512</li>
1513<li>If the expression was computable in any way, this method returns the APValue
1514 for the result of the expression.</li>
1515<li>If the expression is not evaluatable at all, this method returns
1516 information on one of the problems with the expression. This includes a
1517 SourceLocation for where the problem is, and a diagnostic ID that explains
1518 the problem. The diagnostic should be have ERROR type.</li>
1519<li>If the expression is not an integer constant expression, this method returns
1520 information on one of the problems with the expression. This includes a
1521 SourceLocation for where the problem is, and a diagnostic ID that explains
1522 the problem. The diagnostic should be have EXTENSION type.</li>
1523</ul>
1524
1525<p>This information gives various clients the flexibility that they want, and we
1526will eventually have some helper methods for various extensions. For example,
1527Sema should have a <tt>Sema::VerifyIntegerConstantExpression</tt> method, which
1528calls Evaluate on the expression. If the expression is not foldable, the error
1529is emitted, and it would return true. If the expression is not an i-c-e, the
1530EXTENSION diagnostic is emitted. Finally it would return false to indicate that
1531the AST is ok.</p>
1532
1533<p>Other clients can use the information in other ways, for example, codegen can
1534just use expressions that are foldable in any way.</p>
1535
1536<!-- ========== -->
1537<h4>Extensions</h4>
1538<!-- ========== -->
1539
Chris Lattner552de0a2008-11-23 08:16:56 +0000[diff] [blame]1540<p>This section describes how some of the various extensions Clang supports
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1541interacts with constant evaluation:</p>
1542
1543<ul>
1544<li><b><tt>__extension__</tt></b>: The expression form of this extension causes
1545 any evaluatable subexpression to be accepted as an integer constant
1546 expression.</li>
1547<li><b><tt>__builtin_constant_p</tt></b>: This returns true (as a integer
Chris Lattner28daa532008-12-12 06:55:44 +0000[diff] [blame]1548 constant expression) if the operand is any evaluatable constant. As a
1549 special case, if <tt>__builtin_constant_p</tt> is the (potentially
1550 parenthesized) condition of a conditional operator expression ("?:"), only
Chris Lattner42b83dd2008-12-12 18:00:51 +0000[diff] [blame]1551 the true side of the conditional operator is considered, and it is evaluated
1552 with full constant folding.</li>
Chris Lattner7bad1992008-11-16 21:48:07 +0000[diff] [blame]1553<li><b><tt>__builtin_choose_expr</tt></b>: The condition is required to be an
1554 integer constant expression, but we accept any constant as an "extension of
1555 an extension". This only evaluates one operand depending on which way the
1556 condition evaluates.</li>
1557<li><b><tt>__builtin_classify_type</tt></b>: This always returns an integer
1558 constant expression.</li>
1559<li><b><tt>__builtin_inf,nan,..</tt></b>: These are treated just like a
1560 floating-point literal.</li>
1561<li><b><tt>__builtin_abs,copysign,..</tt></b>: These are constant folded as
1562 general constant expressions.</li>
1563</ul>
1564
1565
1566
1567
Ted Kremenek17a295d2008-06-11 06:19:49 +0000[diff] [blame]1568</div>
1569</body>
Douglas Gregor2e1cd422008-11-17 14:58:09 +0000[diff] [blame]1570</html>