📄 regex.cxx
字号:
case '^': { if ( /* If at start of pattern, it's an operator. */ p == pattern + 1 /* If context independent, it's an operator. */ || syntax & RE_CONTEXT_INDEP_ANCHORS /* Otherwise, depends on what's come before. */ || at_begline_loc_p (pattern, p, syntax)) BUF_PUSH (begline); else goto normal_char; } break; case '$': { if ( /* If at end of pattern, it's an operator. */ p == pend /* If context independent, it's an operator. */ || syntax & RE_CONTEXT_INDEP_ANCHORS /* Otherwise, depends on what's next. */ || at_endline_loc_p (p, pend, syntax)) BUF_PUSH (endline); else goto normal_char; } break; case '+': case '?': if ((syntax & RE_BK_PLUS_QM) || (syntax & RE_LIMITED_OPS)) goto normal_char; handle_plus: case '*': /* If there is no previous pattern... */ if (!laststart) { if (syntax & RE_CONTEXT_INVALID_OPS) return REG_BADRPT; else if (!(syntax & RE_CONTEXT_INDEP_OPS)) goto normal_char; } { /* Are we optimizing this jump? */ boolean keep_string_p = false; /* 1 means zero (many) matches is allowed. */ char zero_times_ok = 0, many_times_ok = 0; /* If there is a sequence of repetition chars, collapse it down to just one (the right one). We can't combine interval operators with these because of, e.g., `a{2}*', which should only match an even number of `a's. */ for (;;) { zero_times_ok |= c != '+'; many_times_ok |= c != '?'; if (p == pend) break; PATFETCH (c); if (c == '*' || (!(syntax & RE_BK_PLUS_QM) && (c == '+' || c == '?'))) ; else if (syntax & RE_BK_PLUS_QM && c == '\\') { if (p == pend) return REG_EESCAPE; PATFETCH (c1); if (!(c1 == '+' || c1 == '?')) { PATUNFETCH; PATUNFETCH; break; } c = c1; } else { PATUNFETCH; break; } /* If we get here, we found another repeat character. */ } /* Star, etc. applied to an empty pattern is equivalent to an empty pattern. */ if (!laststart) break; /* Now we know whether or not zero matches is allowed and also whether or not two or more matches is allowed. */ if (many_times_ok) { /* More than one repetition is allowed, so put in at the end a backward relative jump from `b' to before the next jump we're going to put in below (which jumps from laststart to after this jump). But if we are at the `*' in the exact sequence `.*\n', insert an unconditional jump backwards to the ., instead of the beginning of the loop. This way we only push a failure point once, instead of every time through the loop. */ assert (p - 1 > pattern); /* Allocate the space for the jump. */ GET_BUFFER_SPACE (3); /* We know we are not at the first character of the pattern, because laststart was nonzero. And we've already incremented `p', by the way, to be the character after the `*'. Do we have to do something analogous here for null bytes, because of RE_DOT_NOT_NULL? */ if (TRANSLATE (*(p - 2)) == TRANSLATE ('.') && zero_times_ok && p < pend && TRANSLATE (*p) == TRANSLATE ('\n') && !(syntax & RE_DOT_NEWLINE)) { /* We have .*\n. */ STORE_JUMP (jump, b, laststart); keep_string_p = true; } else /* Anything else. */ STORE_JUMP (maybe_pop_jump, b, laststart - 3); /* We've added more stuff to the buffer. */ b += 3; } /* On failure, jump from laststart to b + 3, which will be the end of the buffer after this jump is inserted. */ GET_BUFFER_SPACE (3); INSERT_JUMP (keep_string_p ? on_failure_keep_string_jump : on_failure_jump, laststart, b + 3); pending_exact = 0; b += 3; if (!zero_times_ok) { /* At least one repetition is required, so insert a `dummy_failure_jump' before the initial `on_failure_jump' instruction of the loop. This effects a skip over that instruction the first time we hit that loop. */ GET_BUFFER_SPACE (3); INSERT_JUMP (dummy_failure_jump, laststart, laststart + 6); b += 3; } } break; case '.': laststart = b; BUF_PUSH (anychar); break; case '[': { boolean had_char_class = false; if (p == pend) return REG_EBRACK; /* Ensure that we have enough space to push a charset: the opcode, the length count, and the bitset; 34 bytes in all. */ GET_BUFFER_SPACE (34); laststart = b; /* We test `*p == '^' twice, instead of using an if statement, so we only need one BUF_PUSH. */ BUF_PUSH (*p == '^' ? charset_not : charset); if (*p == '^') p++; /* Remember the first position in the bracket expression. */ p1 = p; /* Push the number of bytes in the bitmap. */ BUF_PUSH ((1 << BYTEWIDTH) / BYTEWIDTH); /* Clear the whole map. */ bzero (b, (1 << BYTEWIDTH) / BYTEWIDTH); /* charset_not matches newline according to a syntax bit. */ if ((re_opcode_t) b[-2] == charset_not && (syntax & RE_HAT_LISTS_NOT_NEWLINE)) SET_LIST_BIT ('\n'); /* Read in characters and ranges, setting map bits. */ for (;;) { if (p == pend) return REG_EBRACK; PATFETCH (c); /* \ might escape characters inside [...] and [^...]. */ if ((syntax & RE_BACKSLASH_ESCAPE_IN_LISTS) && c == '\\') { if (p == pend) return REG_EESCAPE; PATFETCH (c1); SET_LIST_BIT (c1); continue; } /* Could be the end of the bracket expression. If it's not (i.e., when the bracket expression is `[]' so far), the ']' character bit gets set way below. */ if (c == ']' && p != p1 + 1) break; /* Look ahead to see if it's a range when the last thing was a character class. */ if (had_char_class && c == '-' && *p != ']') return REG_ERANGE; /* Look ahead to see if it's a range when the last thing was a character: if this is a hyphen not at the beginning or the end of a list, then it's the range operator. */ if (c == '-' && !(p - 2 >= pattern && p[-2] == '[') && !(p - 3 >= pattern && p[-3] == '[' && p[-2] == '^') && *p != ']') { reg_errcode_t ret = compile_range (&p, pend, translate, syntax, b); if (ret != REG_NOERROR) return ret; } else if (p[0] == '-' && p[1] != ']') { /* This handles ranges made up of characters only. */ reg_errcode_t ret; /* Move past the `-'. */ PATFETCH (c1); ret = compile_range (&p, pend, translate, syntax, b); if (ret != REG_NOERROR) return ret; } /* See if we're at the beginning of a possible character class. */ else if (syntax & RE_CHAR_CLASSES && c == '[' && *p == ':') { /* Leave room for the null. */ char str[CHAR_CLASS_MAX_LENGTH + 1]; PATFETCH (c); c1 = 0; /* If pattern is `[[:'. */ if (p == pend) return REG_EBRACK; for (;;) { PATFETCH (c); if (c == ':' || c == ']' || p == pend || c1 == CHAR_CLASS_MAX_LENGTH) break; str[c1++] = c; } str[c1] = '\0'; /* If isn't a word bracketed by `[:' and:`]': undo the ending character, the letters, and leave the leading `:' and `[' (but set bits for them). */ if (c == ':' && *p == ']') { int ch; boolean is_alnum = STREQ (str, "alnum"); boolean is_alpha = STREQ (str, "alpha"); boolean is_blank = STREQ (str, "blank"); boolean is_cntrl = STREQ (str, "cntrl"); boolean is_digit = STREQ (str, "digit"); boolean is_graph = STREQ (str, "graph"); boolean is_lower = STREQ (str, "lower"); boolean is_print = STREQ (str, "print"); boolean is_punct = STREQ (str, "punct"); boolean is_space = STREQ (str, "space"); boolean is_upper = STREQ (str, "upper"); boolean is_xdigit = STREQ (str, "xdigit"); if (!IS_CHAR_CLASS (str)) return REG_ECTYPE; /* Throw away the ] at the end of the character class. */ PATFETCH (c); if (p == pend) return REG_EBRACK; for (ch = 0; ch < 1 << BYTEWIDTH; ch++) { if ( (is_alnum && ISALNUM (ch)) || (is_alpha && ISALPHA (ch)) || (is_blank && ISBLANK (ch)) || (is_cntrl && ISCNTRL (ch)) || (is_digit && ISDIGIT (ch)) || (is_graph && ISGRAPH (ch)) || (is_lower && ISLOWER (ch)) || (is_print && ISPRINT (ch)) || (is_punct && ISPUNCT (ch)) || (is_space && ISSPACE (ch)) || (is_upper && ISUPPER (ch)) || (is_xdigit && ISXDIGIT (ch))) SET_LIST_BIT (ch); } had_char_class = true; } else { c1++; while (c1--) PATUNFETCH; SET_LIST_BIT ('['); SET_LIST_BIT (':'); had_char_class = false; } } else { had_char_class = false; SET_LIST_BIT (c); } } /* Discard any (non)matching list bytes that are all 0 at the end of the map. Decrease the map-length byte too. */ while ((int) b[-1] > 0 && b[b[-1] - 1] == 0) b[-1]--; b += b[-1]; } break; case '(': if (syntax & RE_NO_BK_PARENS) goto handle_open; else goto normal_char; case ')': if (syntax & RE_NO_BK_PARENS) goto handle_close; else goto normal_char; case '\n': if (syntax & RE_NEWLINE_ALT) goto handle_alt; else goto normal_char; case '|': if (syntax & RE_NO_BK_VBAR) goto handle_alt; else goto normal_char; case '{': if (syntax & RE_INTERVALS && syntax & RE_NO_BK_BRACES) goto handle_interval; else goto normal_char; case '\\': if (p == pend) return REG_EESCAPE; /* Do not translate the character after the \, so that we can distinguish, e.g., \B from \b, even if we normally would translate, e.g., B to b. */ PATFETCH_RAW (c); switch (c) { case '(': if (syntax & RE_NO_BK_PARENS) goto normal_backslash; handle_open: bufp->re_nsub++; regnum++; if (COMPILE_STACK_FULL) { RETALLOC (compile_stack.stack, compile_stack.size << 1, compile_stack_elt_t);
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -