static void readAndEmitTag (const unsigned char** cp, rubyKind expected_kind)
{
if (isspace (**cp))
{
vString *name = vStringNew ();
rubyKind actual_kind = parseIdentifier (cp, name, expected_kind);
if (actual_kind == K_UNDEFINED || vStringLength (name) == 0)
{
/*
* What kind of tags should we create for code like this?
*
* %w(self.clfloor clfloor).each do |name|
* module_eval <<-"end;"
* def #{name}(x, y=1)
* q, r = x.divmod(y)
* q = q.to_i
* return q, r
* end
* end;
* end
*
* Or this?
*
* class << HTTP
*
* For now, we don't create any.
*/
enterUnnamedScope ();
}
else
{
emitRubyTag (name, actual_kind);
}
vStringDelete (name);
}
}
static void findRubyTags (void)
{
const unsigned char *line;
boolean inMultiLineComment = FALSE;
nesting = nestingLevelsNew (0);
/* FIXME: this whole scheme is wrong, because Ruby isn't line-based.
* You could perfectly well write:
*
* def
* method
* puts("hello")
* end
*
* if you wished, and this function would fail to recognize anything.
*/
while ((line = readLineFromInputFile ()) != NULL)
{
const unsigned char *cp = line;
/* if we expect a separator after a while, for, or until statement
* separators are "do", ";" or newline */
boolean expect_separator = FALSE;
if (canMatch (&cp, "=begin", isWhitespace))
{
inMultiLineComment = TRUE;
continue;
}
if (canMatch (&cp, "=end", isWhitespace))
{
inMultiLineComment = FALSE;
continue;
}
if (inMultiLineComment)
continue;
skipWhitespace (&cp);
/* Avoid mistakenly starting a scope for modifiers such as
*
* return if <exp>
*
* FIXME: this is fooled by code such as
*
* result = if <exp>
* <a>
* else
* <b>
* end
*
* FIXME: we're also fooled if someone does something heinous such as
*
* puts("hello") \
* unless <exp>
*/
if (canMatchKeyword (&cp, "for") ||
canMatchKeyword (&cp, "until") ||
canMatchKeyword (&cp, "while"))
{
expect_separator = TRUE;
enterUnnamedScope ();
}
else if (canMatchKeyword (&cp, "case") ||
canMatchKeyword (&cp, "if") ||
canMatchKeyword (&cp, "unless"))
{
enterUnnamedScope ();
}
/*
* "module M", "class C" and "def m" should only be at the beginning
* of a line.
*/
if (canMatchKeyword (&cp, "module"))
{
readAndEmitTag (&cp, K_MODULE);
}
else if (canMatchKeyword (&cp, "class"))
{
readAndEmitTag (&cp, K_CLASS);
}
else if (canMatchKeyword (&cp, "def"))
{
rubyKind kind = K_METHOD;
NestingLevel *nl = nestingLevelsGetCurrent (nesting);
tagEntryInfo *e = getEntryOfNestingLevel (nl);
/* if the def is inside an unnamed scope at the class level, assume
* it's from a singleton from a construct like this:
*
* class C
* class << self
* def singleton
* ...
* end
* end
* end
*/
if (e && (e->kind - RubyKinds) == K_CLASS && strlen (e->name) == 0)
kind = K_SINGLETON;
readAndEmitTag (&cp, kind);
}
while (*cp != '\0')
{
/* FIXME: we don't cope with here documents,
* or regular expression literals, or ... you get the idea.
* Hopefully, the restriction above that insists on seeing
* definitions at the starts of lines should keep us out of
* mischief.
*/
if (inMultiLineComment || isspace (*cp))
{
++cp;
}
else if (*cp == '#')
{
/* FIXME: this is wrong, but there *probably* won't be a
* definition after an interpolated string (where # doesn't
* mean 'comment').
*/
break;
}
else if (canMatchKeyword (&cp, "begin"))
{
enterUnnamedScope ();
}
else if (canMatchKeyword (&cp, "do"))
{
if (! expect_separator)
enterUnnamedScope ();
else
expect_separator = FALSE;
}
else if (canMatchKeyword (&cp, "end") && nesting->n > 0)
{
/* Leave the most recent scope. */
nestingLevelsPop (nesting);
}
else if (*cp == '"')
{
/* Skip string literals.
* FIXME: should cope with escapes and interpolation.
*/
do {
++cp;
} while (*cp != 0 && *cp != '"');
if (*cp == '"')
cp++; /* skip the last found '"' */
}
else if (*cp == ';')
{
++cp;
expect_separator = FALSE;
}
else if (*cp != '\0')
{
do
++cp;
while (isIdentChar (*cp));
}
}
}
nestingLevelsFree (nesting);
}
static void findRubyTags (void)
{
const unsigned char *line;
boolean inMultiLineComment = FALSE;
nesting = stringListNew ();
/* FIXME: this whole scheme is wrong, because Ruby isn't line-based.
* You could perfectly well write:
*
* def
* method
* puts("hello")
* end
*
* if you wished, and this function would fail to recognize anything.
*/
while ((line = fileReadLine ()) != NULL)
{
const unsigned char *cp = line;
if (canMatch (&cp, "=begin"))
{
inMultiLineComment = TRUE;
continue;
}
if (canMatch (&cp, "=end"))
{
inMultiLineComment = FALSE;
continue;
}
skipWhitespace (&cp);
/* Avoid mistakenly starting a scope for modifiers such as
*
* return if <exp>
*
* FIXME: this is fooled by code such as
*
* result = if <exp>
* <a>
* else
* <b>
* end
*
* FIXME: we're also fooled if someone does something heinous such as
*
* puts("hello") \
* unless <exp>
*/
if (canMatch (&cp, "case") || canMatch (&cp, "for") ||
canMatch (&cp, "if") || canMatch (&cp, "unless") ||
canMatch (&cp, "while"))
{
enterUnnamedScope ();
}
/*
* "module M", "class C" and "def m" should only be at the beginning
* of a line.
*/
if (canMatch (&cp, "module"))
{
readAndEmitTag (&cp, K_MODULE);
}
else if (canMatch (&cp, "class"))
{
readAndEmitTag (&cp, K_CLASS);
}
else if (canMatch (&cp, "def"))
{
readAndEmitTag (&cp, K_METHOD);
}
while (*cp != '\0')
{
/* FIXME: we don't cope with here documents,
* or regular expression literals, or ... you get the idea.
* Hopefully, the restriction above that insists on seeing
* definitions at the starts of lines should keep us out of
* mischief.
*/
if (inMultiLineComment || isspace (*cp))
{
++cp;
}
else if (*cp == '#')
{
/* FIXME: this is wrong, but there *probably* won't be a
* definition after an interpolated string (where # doesn't
* mean 'comment').
*/
break;
}
else if (canMatch (&cp, "begin") || canMatch (&cp, "do"))
{
enterUnnamedScope ();
}
else if (canMatch (&cp, "end") && stringListCount (nesting) > 0)
{
/* Leave the most recent scope. */
vStringDelete (stringListLast (nesting));
stringListRemoveLast (nesting);
}
else if (*cp == '"')
{
/* Skip string literals.
* FIXME: should cope with escapes and interpolation.
*/
do {
++cp;
} while (*cp != 0 && *cp != '"');
}
else if (*cp != '\0')
{
do
++cp;
while (isalnum (*cp) || *cp == '_');
}
}
}
stringListDelete (nesting);
}
