fixed a bug in handling a regular expression starting with a backslash.

- a regular expression like /\// could not be handled properly without 
this fix
This commit is contained in:
2009-03-26 08:53:10 +00:00
parent ce46d8f641
commit ba8bd06016
4 changed files with 79 additions and 27 deletions

View File

@ -1,5 +1,5 @@
/*
* $Id: parse.c 85 2009-02-26 10:56:12Z hyunghwan.chung $
* $Id: parse.c 113 2009-03-25 14:53:10Z hyunghwan.chung $
*
Copyright 2006-2009 Chung, Hyung-Hwan.
@ -220,7 +220,7 @@ static int get_charstr (qse_awk_t* awk);
static int get_rexstr (qse_awk_t* awk);
static int get_string (
qse_awk_t* awk, qse_char_t end_char,
qse_char_t esc_char, qse_bool_t keep_esc_char);
qse_char_t esc_char, qse_bool_t keep_esc_char, int preescaped);
static int get_char (qse_awk_t* awk);
static int unget_char (qse_awk_t* awk, qse_cint_t c);
static int skip_spaces (qse_awk_t* awk);
@ -2930,7 +2930,9 @@ static qse_awk_nde_t* parse_primary (qse_awk_t* awk, qse_size_t line)
int errnum;
/* the regular expression is tokenized here because
* of the context-sensitivity of the slash symbol */
* of the context-sensitivity of the slash symbol.
* if TOKEN_DIV is seen as a primary, it tries to compile
* it as a regular expression */
SET_TOKEN_TYPE (awk, TOKEN_REX);
qse_str_clear (awk->token.name);
@ -4567,7 +4569,6 @@ static qse_awk_nde_t* parse_print (qse_awk_t* awk, qse_size_t line, int type)
return (qse_awk_nde_t*)nde;
}
static int get_token (qse_awk_t* awk)
{
qse_cint_t c;
@ -5091,7 +5092,7 @@ static int get_charstr (qse_awk_t* awk)
* has been called */
ADD_TOKEN_CHAR (awk, awk->src.lex.curc);
}
return get_string (awk, QSE_T('\"'), QSE_T('\\'), QSE_FALSE);
return get_string (awk, QSE_T('\"'), QSE_T('\\'), QSE_FALSE, 0);
}
static int get_rexstr (qse_awk_t* awk)
@ -5099,23 +5100,44 @@ static int get_rexstr (qse_awk_t* awk)
if (awk->src.lex.curc == QSE_T('/'))
{
/* this part of the function is different from get_charstr
* because of the way this function is called */
* because of the way this function is called.
* this condition is met when the input is //.
* the first / has been tokenized to TOKEN_DIV already.
* if TOKEN_DIV is seen as a primary, this function is called.
* as the token buffer has been cleared by the caller and
* the token type is set to TOKEN_REX, this function can
* just return after reading the next character */
GET_CHAR (awk);
return 0;
}
else
{
ADD_TOKEN_CHAR (awk, awk->src.lex.curc);
return get_string (awk, QSE_T('/'), QSE_T('\\'), QSE_TRUE);
int escaped = 0;
if (awk->src.lex.curc == QSE_T('\\'))
{
/* for input like /\//, this condition is met.
* the initial escape character is added when the
* second charater is handled in get_string() */
escaped = 1;
}
else
{
/* add other initial characters here as get_string()
* begins with reading the next character */
ADD_TOKEN_CHAR (awk, awk->src.lex.curc);
}
return get_string (awk,
QSE_T('/'), QSE_T('\\'), QSE_TRUE, escaped);
}
}
static int get_string (
qse_awk_t* awk, qse_char_t end_char,
qse_char_t esc_char, qse_bool_t keep_esc_char)
qse_char_t esc_char, qse_bool_t keep_esc_char,
int preescaped)
{
qse_cint_t c;
int escaped = 0;
int escaped = preescaped;
int digit_count = 0;
qse_cint_t c_acc = 0;