ruby.c

   1 /*
   2 *   $Id: ruby.c 571 2007-06-24 23:32:14Z elliotth $
   3 *
   4 *   Copyright (c) 2000-2001, Thaddeus Covert <sahuagin@mediaone.net>
   5 *   Copyright (c) 2002 Matthias Veit <matthias_veit@yahoo.de>
   6 *   Copyright (c) 2004 Elliott Hughes <enh@acm.org>
   7 *
   8 *   This source code is released for free distribution under the terms of the
   9 *   GNU General Public License.
  10 *
  11 *   This module contains functions for generating tags for Ruby language
  12 *   files.
  13 */
  14
  15 /*
  16 *   INCLUDE FILES
  17 */
  18 #include "general.h"  /* must always come first */
  19
  20 #include <string.h>
  21
  22 #include "entry.h"
  23 #include "parse.h"
  24 #include "read.h"
  25 #include "vstring.h"
  26
  27 /*
  28 *   DATA DECLARATIONS
  29 */
  30 typedef enum {
  31         K_UNDEFINED = -1, K_CLASS, K_METHOD, K_MODULE, K_SINGLETON
  32 } rubyKind;
  33
  34 /*
  35 *   DATA DEFINITIONS
  36 */
  37 static kindOption RubyKinds [] = {
  38         { TRUE, 'c', "class",  "classes" },
  39         { TRUE, 'f', "method", "methods" },
  40         { TRUE, 'm', "module", "modules" },
  41         { TRUE, 'F', "singleton method", "singleton methods" }
  42 };
  43
  44 static stringList* nesting = 0;
  45
  46 /*
  47 *   FUNCTION DEFINITIONS
  48 */
  49
  50 /*
  51 * Returns a string describing the scope in 'list'.
  52 * We record the current scope as a list of entered scopes.
  53 * Scopes corresponding to 'if' statements and the like are
  54 * represented by empty strings. Scopes corresponding to
  55 * modules and classes are represented by the name of the
  56 * module or class.
  57 */
  58 static vString* stringListToScope (const stringList* list)
  59 {
  60         unsigned int i;
  61         unsigned int chunks_output = 0;
  62         vString* result = vStringNew ();
  63         const unsigned int max = stringListCount (list);
  64         for (i = 0; i < max; ++i)
  65         {
  66             vString* chunk = stringListItem (list, i);
  67             if (vStringLength (chunk) > 0)
  68             {
  69                 vStringCatS (result, (chunks_output++ > 0) ? "." : "");
  70                 vStringCatS (result, vStringValue (chunk));
  71             }
  72         }
  73         return result;
  74 }
  75
  76 /*
  77 * Attempts to advance 's' past 'literal'.
  78 * Returns TRUE if it did, FALSE (and leaves 's' where
  79 * it was) otherwise.
  80 */
  81 static boolean canMatch (const unsigned char** s, const char* literal)
  82 {
  83         const int literal_length = strlen (literal);
  84         const unsigned char next_char = *(*s + literal_length);
  85         if (strncmp ((const char*) *s, literal, literal_length) != 0)
  86         {
  87             return FALSE;
  88         }
  89         /* Additionally check that we're at the end of a token. */
  90         if ( ! (next_char == 0 || isspace (next_char) || next_char == '('))
  91         {
  92             return FALSE;
  93         }
  94         *s += literal_length;
  95         return TRUE;
  96 }
  97
  98 /*
  99 * Attempts to advance 'cp' past a Ruby operator method name. Returns
 100 * TRUE if successful (and copies the name into 'name'), FALSE otherwise.
 101 */
 102 static boolean parseRubyOperator (vString* name, const unsigned char** cp)
 103 {
 104         static const char* RUBY_OPERATORS[] = {
 105             "[]", "[]=",
 106             "**",
 107             "!", "~", "+@", "-@",
 108             "*", "/", "%",
 109             "+", "-",
 110             ">>", "<<",
 111             "&",
 112             "^", "|",
 113             "<=", "<", ">", ">=",
 114             "<=>", "==", "===", "!=", "=~", "!~",
 115             "`",
 116             0
 117         };
 118         int i;
 119         for (i = 0; RUBY_OPERATORS[i] != 0; ++i)
 120         {
 121             if (canMatch (cp, RUBY_OPERATORS[i]))
 122             {
 123                 vStringCatS (name, RUBY_OPERATORS[i]);
 124                 return TRUE;
 125             }
 126         }
 127         return FALSE;
 128 }
 129
 130 /*
 131 * Emits a tag for the given 'name' of kind 'kind' at the current nesting.
 132 */
 133 static void emitRubyTag (vString* name, rubyKind kind)
 134 {
 135         tagEntryInfo tag;
 136         vString* scope;
 137
 138         vStringTerminate (name);
 139         scope = stringListToScope (nesting);
 140
 141         initTagEntry (&tag, vStringValue (name));
 142         if (vStringLength (scope) > 0) {
 143             tag.extensionFields.scope [0] = "class";
 144             tag.extensionFields.scope [1] = vStringValue (scope);
 145         }
 146         tag.kindName = RubyKinds [kind].name;
 147         tag.kind = RubyKinds [kind].letter;
 148         makeTagEntry (&tag);
 149
 150         stringListAdd (nesting, vStringNewCopy (name));
 151
 152         vStringClear (name);
 153         vStringDelete (scope);
 154 }
 155
 156 /* Tests whether 'ch' is a character in 'list'. */
 157 static boolean charIsIn (char ch, const char* list)
 158 {
 159         return (strchr (list, ch) != 0);
 160 }
 161
 162 /* Advances 'cp' over leading whitespace. */
 163 static void skipWhitespace (const unsigned char** cp)
 164 {
 165         while (isspace (**cp))
 166         {
 167             ++*cp;
 168         }
 169 }
 170
 171 /*
 172 * Copies the characters forming an identifier from *cp into
 173 * name, leaving *cp pointing to the character after the identifier.
 174 */
 175 static rubyKind parseIdentifier (
 176                 const unsigned char** cp, vString* name, rubyKind kind)
 177 {
 178         /* Method names are slightly different to class and variable names.
 179          * A method name may optionally end with a question mark, exclamation
 180          * point or equals sign. These are all part of the name.
 181          * A method name may also contain a period if it's a singleton method.
 182          */
 183         const char* also_ok = (kind == K_METHOD) ? "_.?!=" : "_";
 184
 185         skipWhitespace (cp);
 186
 187         /* Check for an anonymous (singleton) class such as "class << HTTP". */
 188         if (kind == K_CLASS && **cp == '<' && *(*cp + 1) == '<')
 189         {
 190                 return K_UNDEFINED;
 191         }
 192
 193         /* Check for operators such as "def []=(key, val)". */
 194         if (kind == K_METHOD || kind == K_SINGLETON)
 195         {
 196                 if (parseRubyOperator (name, cp))
 197                 {
 198                         return kind;
 199                 }
 200         }
 201
 202         /* Copy the identifier into 'name'. */
 203         while (**cp != 0 && (isalnum (**cp) || charIsIn (**cp, also_ok)))
 204         {
 205                 char last_char = **cp;
 206
 207                 vStringPut (name, last_char);
 208                 ++*cp;
 209
 210                 if (kind == K_METHOD)
 211                 {
 212                         /* Recognize singleton methods. */
 213                         if (last_char == '.')
 214                         {
 215                                 vStringTerminate (name);
 216                                 vStringClear (name);
 217                                 return parseIdentifier (cp, name, K_SINGLETON);
 218                         }
 219
 220                         /* Recognize characters which mark the end of a method name. */
 221                         if (charIsIn (last_char, "?!="))
 222                         {
 223                                 break;
 224                         }
 225                 }
 226         }
 227         return kind;
 228 }
 229
 230 static void readAndEmitTag (const unsigned char** cp, rubyKind expected_kind)
 231 {
 232         if (isspace (**cp))
 233         {
 234                 vString *name = vStringNew ();
 235                 rubyKind actual_kind = parseIdentifier (cp, name, expected_kind);
 236
 237                 if (actual_kind == K_UNDEFINED || vStringLength (name) == 0)
 238                 {
 239                         /*
 240                         * What kind of tags should we create for code like this?
 241                         *
 242                         *    %w(self.clfloor clfloor).each do |name|
 243                         *        module_eval <<-"end;"
 244                         *            def #{name}(x, y=1)
 245                         *                q, r = x.divmod(y)
 246                         *                q = q.to_i
 247                         *                return q, r
 248                         *            end
 249                         *        end;
 250                         *    end
 251                         *
 252                         * Or this?
 253                         *
 254                         *    class << HTTP
 255                         *
 256                         * For now, we don't create any.
 257                         */
 258                 }
 259                 else
 260                 {
 261                         emitRubyTag (name, actual_kind);
 262                 }
 263                 vStringDelete (name);
 264         }
 265 }
 266
 267 static void enterUnnamedScope (void)
 268 {
 269         stringListAdd (nesting, vStringNewInit (""));
 270 }
 271
 272 static void findRubyTags (void)
 273 {
 274         const unsigned char *line;
 275         boolean inMultiLineComment = FALSE;
 276
 277         nesting = stringListNew ();
 278
 279         /* FIXME: this whole scheme is wrong, because Ruby isn't line-based.
 280         * You could perfectly well write:
 281         *
 282         *  def
 283         *  method
 284         *   puts("hello")
 285         *  end
 286         *
 287         * if you wished, and this function would fail to recognize anything.
 288         */
 289         while ((line = fileReadLine ()) != NULL)
 290         {
 291                 const unsigned char *cp = line;
 292
 293                 if (canMatch (&cp, "=begin"))
 294                 {
 295                         inMultiLineComment = TRUE;
 296                         continue;
 297                 }
 298                 if (canMatch (&cp, "=end"))
 299                 {
 300                         inMultiLineComment = FALSE;
 301                         continue;
 302                 }
 303
 304                 skipWhitespace (&cp);
 305
 306                 /* Avoid mistakenly starting a scope for modifiers such as
 307                 *
 308                 *   return if <exp>
 309                 *
 310                 * FIXME: this is fooled by code such as
 311                 *
 312                 *   result = if <exp>
 313                 *               <a>
 314                 *            else
 315                 *               <b>
 316                 *            end
 317                 *
 318                 * FIXME: we're also fooled if someone does something heinous such as
 319                 *
 320                 *   puts("hello") \
 321                 *       unless <exp>
 322                 */
 323                 if (canMatch (&cp, "case") || canMatch (&cp, "for") ||
 324                         canMatch (&cp, "if") || canMatch (&cp, "unless") ||
 325                         canMatch (&cp, "while"))
 326                 {
 327                         enterUnnamedScope ();
 328                 }
 329
 330                 /*
 331                 * "module M", "class C" and "def m" should only be at the beginning
 332                 * of a line.
 333                 */
 334                 if (canMatch (&cp, "module"))
 335                 {
 336                         readAndEmitTag (&cp, K_MODULE);
 337                 }
 338                 else if (canMatch (&cp, "class"))
 339                 {
 340                         readAndEmitTag (&cp, K_CLASS);
 341                 }
 342                 else if (canMatch (&cp, "def"))
 343                 {
 344                         readAndEmitTag (&cp, K_METHOD);
 345                 }
 346
 347                 while (*cp != '\0')
 348                 {
 349                         /* FIXME: we don't cope with here documents,
 350                         * or regular expression literals, or ... you get the idea.
 351                         * Hopefully, the restriction above that insists on seeing
 352                         * definitions at the starts of lines should keep us out of
 353                         * mischief.
 354                         */
 355                         if (inMultiLineComment || isspace (*cp))
 356                         {
 357                                 ++cp;
 358                         }
 359                         else if (*cp == '#')
 360                         {
 361                                 /* FIXME: this is wrong, but there *probably* won't be a
 362                                 * definition after an interpolated string (where # doesn't
 363                                 * mean 'comment').
 364                                 */
 365                                 break;
 366                         }
 367                         else if (canMatch (&cp, "begin") || canMatch (&cp, "do"))
 368                         {
 369                                 enterUnnamedScope ();
 370                         }
 371                         else if (canMatch (&cp, "end") && stringListCount (nesting) > 0)
 372                         {
 373                                 /* Leave the most recent scope. */
 374                                 vStringDelete (stringListLast (nesting));
 375                                 stringListRemoveLast (nesting);
 376                         }
 377                         else if (*cp == '"')
 378                         {
 379                                 /* Skip string literals.
 380                                  * FIXME: should cope with escapes and interpolation.
 381                                  */
 382                                 do {
 383                                         ++cp;
 384                                 } while (*cp != 0 && *cp != '"');
 385                         }
 386                         else if (*cp != '\0')
 387                         {
 388                                 do
 389                                         ++cp;
 390                                 while (isalnum (*cp) || *cp == '_');
 391                         }
 392                 }
 393         }
 394         stringListDelete (nesting);
 395 }
 396
 397 extern parserDefinition* RubyParser (void)
 398 {
 399         static const char *const extensions [] = { "rb", "ruby", NULL };
 400         parserDefinition* def = parserNew ("Ruby");
 401         def->kinds      = RubyKinds;
 402         def->kindCount  = KIND_COUNT (RubyKinds);
 403         def->extensions = extensions;
 404         def->parser     = findRubyTags;
 405         return def;
 406 }
 407
 408 /* vi:set tabstop=4 shiftwidth=4: */