home *** CD-ROM | disk | FTP | other *** search
Text File | 1991-03-05 | 37.0 KB | 1,267 lines |
-
-
-
- /* Like re_search_2, below, but only one string is specified, and
- doesn't let you say where to stop matching. */
-
- int
- re_search (pbufp, string, size, startpos, range, regs)
- struct re_pattern_buffer *pbufp;
- char *string;
- int size, startpos, range;
- struct re_registers *regs;
- {
- return re_search_2 (pbufp, (char *) 0, 0, string, size, startpos, range,
- regs, size);
- }
-
-
- /* Using the compiled pattern in PBUFP->buffer, first tries to match the
- virtual concatenation of STRING1 and STRING2, starting first at index
- STARTPOS, then at STARTPOS + 1, and so on. RANGE is the number of
- places to try before giving up. If RANGE is negative, it searches
- backwards, i.e., the starting positions tried are STARTPOS, STARTPOS
- - 1, etc. STRING1 and STRING2 are of SIZE1 and SIZE2, respectively.
- In REGS, return the indices of the virtual concatenation of STRING1
- and STRING2 that matched the entire PBUFP->buffer and its contained
- subexpressions. Do not consider matching one past the index MSTOP in
- the virtual concatenation of STRING1 and STRING2.
-
- The value returned is the position in the strings at which the match
- was found, or -1 if no match was found, or -2 if error (such as
- failure stack overflow). */
-
- int
- re_search_2 (pbufp, string1, size1, string2, size2, startpos, range,
- regs, mstop)
- struct re_pattern_buffer *pbufp;
- char *string1, *string2;
- int size1, size2;
- int startpos;
- register int range;
- struct re_registers *regs;
- int mstop;
- {
- register char *fastmap = pbufp->fastmap;
- register unsigned char *translate = (unsigned char *) pbufp->translate;
- int total_size = size1 + size2;
- int endpos = startpos + range;
- int val;
-
- /* Check for out-of-range starting position. */
- if (startpos < 0 || startpos > total_size)
- return -1;
-
- /* Fix up range if it would eventually take startpos outside of the
- virtual concatenation of string1 and string2. */
- if (endpos < -1)
- range = -1 - startpos;
- else if (endpos > total_size)
- range = total_size - startpos;
-
- /* Update the fastmap now if not correct already. */
- if (fastmap && !pbufp->fastmap_accurate)
- re_compile_fastmap (pbufp);
-
- /* If the search isn't to be a backwards one, don't waste time in a
- long search for a pattern that says it is anchored. */
- if (pbufp->used > 0 && (enum regexpcode) pbufp->buffer[0] == begbuf
- && range > 0)
- {
- if (startpos > 0)
- return -1;
- else
- range = 1;
- }
-
- while (1)
- {
- /* If a fastmap is supplied, skip quickly over characters that
- cannot possibly be the start of a match. Note, however, that
- if the pattern can possibly match the null string, we must
- test it at each starting point so that we take the first null
- string we get. */
-
- if (fastmap && startpos < total_size && pbufp->can_be_null != 1)
- {
- if (range > 0) /* Searching forwards. */
- {
- register int lim = 0;
- register unsigned char *p;
- int irange = range;
- if (startpos < size1 && startpos + range >= size1)
- lim = range - (size1 - startpos);
-
- p = ((unsigned char *)
- &(startpos >= size1 ? string2 - size1 : string1)[startpos]);
-
- while (range > lim && !fastmap[translate
- ? translate[*p++]
- : *p++])
- range--;
- startpos += irange - range;
- }
- else /* Searching backwards. */
- {
- register unsigned char c;
-
- if (string1 == 0 || startpos >= size1)
- c = string2[startpos - size1];
- else
- c = string1[startpos];
-
- c &= 0xff;
- if (translate ? !fastmap[translate[c]] : !fastmap[c])
- goto advance;
- }
- }
-
- if (range >= 0 && startpos == total_size
- && fastmap && pbufp->can_be_null == 0)
- return -1;
-
- val = re_match_2 (pbufp, string1, size1, string2, size2, startpos,
- regs, mstop);
- if (val >= 0)
- return startpos;
- if (val == -2)
- return -2;
-
- #ifdef C_ALLOCA
- alloca (0);
- #endif /* C_ALLOCA */
-
- advance:
- if (!range)
- break;
- else if (range > 0)
- {
- range--;
- startpos++;
- }
- else
- {
- range++;
- startpos--;
- }
- }
- return -1;
- }
-
-
-
- #ifndef emacs /* emacs never uses this. */
- int
- re_match (pbufp, string, size, pos, regs)
- struct re_pattern_buffer *pbufp;
- char *string;
- int size, pos;
- struct re_registers *regs;
- {
- return re_match_2 (pbufp, (char *) 0, 0, string, size, pos, regs, size);
- }
- #endif /* not emacs */
-
-
- /* The following are used for re_match_2, defined below: */
-
- /* Roughly the maximum number of failure points on the stack. Would be
- exactly that if always pushed MAX_NUM_FAILURE_ITEMS each time we failed. */
-
- int re_max_failures = 2000;
-
- /* Routine used by re_match_2. */
- static int bcmp_translate ();
-
-
- /* Structure and accessing macros used in re_match_2: */
-
- struct register_info
- {
- unsigned is_active : 1;
- unsigned matched_something : 1;
- };
-
- #define IS_ACTIVE(R) ((R).is_active)
- #define MATCHED_SOMETHING(R) ((R).matched_something)
-
-
- /* Macros used by re_match_2: */
-
-
- /* I.e., regstart, regend, and reg_info. */
-
- #define NUM_REG_ITEMS 3
-
- /* We push at most this many things on the stack whenever we
- fail. The `+ 2' refers to PATTERN_PLACE and STRING_PLACE, which are
- arguments to the PUSH_FAILURE_POINT macro. */
-
- #define MAX_NUM_FAILURE_ITEMS (RE_NREGS * NUM_REG_ITEMS + 2)
-
-
- /* We push this many things on the stack whenever we fail. */
-
- #define NUM_FAILURE_ITEMS (last_used_reg * NUM_REG_ITEMS + 2)
-
-
- /* This pushes most of the information about the current state we will want
- if we ever fail back to it. */
-
- #define PUSH_FAILURE_POINT(pattern_place, string_place) \
- { \
- short last_used_reg, this_reg; \
- \
- /* Find out how many registers are active or have been matched. \
- (Aside from register zero, which is only set at the end.) */ \
- for (last_used_reg = RE_NREGS - 1; last_used_reg > 0; last_used_reg--)\
- if (regstart[last_used_reg] != (unsigned char *) -1) \
- break; \
- \
- if (stacke - stackp < NUM_FAILURE_ITEMS) \
- { \
- unsigned char **stackx; \
- if (stacke - stackb > re_max_failures * MAX_NUM_FAILURE_ITEMS) \
- return -2; \
- \
- /* Roughly double the size of the stack. */ \
- stackx = (unsigned char **) alloca (2 * MAX_NUM_FAILURE_ITEMS \
- * (stacke - stackb) \
- * sizeof (unsigned char *));\
- /* Only copy what is in use. */ \
- bcopy (stackb, stackx, (stackp - stackb) * sizeof (char *)); \
- stackp = stackx + (stackp - stackb); \
- stackb = stackx; \
- stacke = stackb + 2 * MAX_NUM_FAILURE_ITEMS * (stacke - stackb);\
- } \
- \
- /* Now push the info for each of those registers. */ \
- for (this_reg = 1; this_reg <= last_used_reg; this_reg++) \
- { \
- *stackp++ = regstart[this_reg]; \
- *stackp++ = regend[this_reg]; \
- *stackp++ = (unsigned char *) ®_info[this_reg]; \
- } \
- \
- /* Push how many registers we saved. */ \
- *stackp++ = (unsigned char *) last_used_reg; \
- \
- *stackp++ = pattern_place; \
- *stackp++ = string_place; \
- }
-
-
- /* This pops what PUSH_FAILURE_POINT pushes. */
-
- #define POP_FAILURE_POINT() \
- { \
- int temp; \
- stackp -= 2; /* Remove failure points. */ \
- temp = (int) *--stackp; /* How many regs pushed. */ \
- temp *= NUM_REG_ITEMS; /* How much to take off the stack. */ \
- stackp -= temp; /* Remove the register info. */ \
- }
-
-
- #define MATCHING_IN_FIRST_STRING (dend == end_match_1)
-
- /* Is true if there is a first string and if PTR is pointing anywhere
- inside it or just past the end. */
-
- #define IS_IN_FIRST_STRING(ptr) \
- (size1 && string1 <= (ptr) && (ptr) <= string1 + size1)
-
- /* Call before fetching a character with *d. This switches over to
- string2 if necessary. */
-
- #define PREFETCH \
- while (d == dend) \
- { \
- /* end of string2 => fail. */ \
- if (dend == end_match_2) \
- goto fail; \
- /* end of string1 => advance to string2. */ \
- d = string2; \
- dend = end_match_2; \
- }
-
-
- /* Call this when have matched something; it sets `matched' flags for the
- registers corresponding to the subexpressions of which we currently
- are inside. */
- #define SET_REGS_MATCHED \
- { unsigned this_reg; \
- for (this_reg = 0; this_reg < RE_NREGS; this_reg++) \
- { \
- if (IS_ACTIVE(reg_info[this_reg])) \
- MATCHED_SOMETHING(reg_info[this_reg]) = 1; \
- else \
- MATCHED_SOMETHING(reg_info[this_reg]) = 0; \
- } \
- }
-
- /* Test if at very beginning or at very end of the virtual concatenation
- of string1 and string2. If there is only one string, we've put it in
- string2. */
-
- #define AT_STRINGS_BEG (d == (size1 ? string1 : string2) || !size2)
- #define AT_STRINGS_END (d == end2)
-
- #define AT_WORD_BOUNDARY \
- (AT_STRINGS_BEG || AT_STRINGS_END || IS_A_LETTER (d - 1) != IS_A_LETTER (d))
-
- /* We have two special cases to check for:
- 1) if we're past the end of string1, we have to look at the first
- character in string2;
- 2) if we're before the beginning of string2, we have to look at the
- last character in string1; we assume there is a string1, so use
- this in conjunction with AT_STRINGS_BEG. */
- #define IS_A_LETTER(d) \
- (SYNTAX ((d) == end1 ? *string2 : (d) == string2 - 1 ? *(end1 - 1) : *(d))\
- == Sword)
-
-
- /* Match the pattern described by PBUFP against the virtual
- concatenation of STRING1 and STRING2, which are of SIZE1 and SIZE2,
- respectively. Start the match at index POS in the virtual
- concatenation of STRING1 and STRING2. In REGS, return the indices of
- the virtual concatenation of STRING1 and STRING2 that matched the
- entire PBUFP->buffer and its contained subexpressions. Do not
- consider matching one past the index MSTOP in the virtual
- concatenation of STRING1 and STRING2.
-
- If pbufp->fastmap is nonzero, then it had better be up to date.
-
- The reason that the data to match are specified as two components
- which are to be regarded as concatenated is so this function can be
- used directly on the contents of an Emacs buffer.
-
- -1 is returned if there is no match. -2 is returned if there is an
- error (such as match stack overflow). Otherwise the value is the
- length of the substring which was matched. */
-
- int
- re_match_2 (pbufp, string1_arg, size1, string2_arg, size2, pos, regs, mstop)
- struct re_pattern_buffer *pbufp;
- char *string1_arg, *string2_arg;
- int size1, size2;
- int pos;
- struct re_registers *regs;
- int mstop;
- {
- register unsigned char *p = (unsigned char *) pbufp->buffer;
-
- /* Pointer to beyond end of buffer. */
- register unsigned char *pend = p + pbufp->used;
-
- unsigned char *string1 = (unsigned char *) string1_arg;
- unsigned char *string2 = (unsigned char *) string2_arg;
- unsigned char *end1; /* Just past end of first string. */
- unsigned char *end2; /* Just past end of second string. */
-
- /* Pointers into string1 and string2, just past the last characters in
- each to consider matching. */
- unsigned char *end_match_1, *end_match_2;
-
- register unsigned char *d, *dend;
- register int mcnt; /* Multipurpose. */
- unsigned char *translate = (unsigned char *) pbufp->translate;
- unsigned is_a_jump_n = 0;
-
- /* Failure point stack. Each place that can handle a failure further
- down the line pushes a failure point on this stack. It consists of
- restart, regend, and reg_info for all registers corresponding to the
- subexpressions we're currently inside, plus the number of such
- registers, and, finally, two char *'s. The first char * is where to
- resume scanning the pattern; the second one is where to resume
- scanning the strings. If the latter is zero, the failure point is a
- ``dummy''; if a failure happens and the failure point is a dummy, it
- gets discarded and the next next one is tried. */
-
- unsigned char *initial_stack[MAX_NUM_FAILURE_ITEMS * NFAILURES];
- unsigned char **stackb = initial_stack;
- unsigned char **stackp = stackb;
- unsigned char **stacke = &stackb[MAX_NUM_FAILURE_ITEMS * NFAILURES];
-
-
- /* Information on the contents of registers. These are pointers into
- the input strings; they record just what was matched (on this
- attempt) by a subexpression part of the pattern, that is, the
- regnum-th regstart pointer points to where in the pattern we began
- matching and the regnum-th regend points to right after where we
- stopped matching the regnum-th subexpression. (The zeroth register
- keeps track of what the whole pattern matches.) */
-
- unsigned char *regstart[RE_NREGS];
- unsigned char *regend[RE_NREGS];
-
- /* The is_active field of reg_info helps us keep track of which (possibly
- nested) subexpressions we are currently in. The matched_something
- field of reg_info[reg_num] helps us tell whether or not we have
- matched any of the pattern so far this time through the reg_num-th
- subexpression. These two fields get reset each time through any
- loop their register is in. */
-
- struct register_info reg_info[RE_NREGS];
-
-
- /* The following record the register info as found in the above
- variables when we find a match better than any we've seen before.
- This happens as we backtrack through the failure points, which in
- turn happens only if we have not yet matched the entire string. */
-
- unsigned best_regs_set = 0;
- unsigned char *best_regstart[RE_NREGS];
- unsigned char *best_regend[RE_NREGS];
-
-
- /* Initialize subexpression text positions to -1 to mark ones that no
- \( or ( and \) or ) has been seen for. Also set all registers to
- inactive and mark them as not having matched anything or ever
- failed. */
- for (mcnt = 0; mcnt < RE_NREGS; mcnt++)
- {
- regstart[mcnt] = regend[mcnt] = (unsigned char *) -1;
- IS_ACTIVE (reg_info[mcnt]) = 0;
- MATCHED_SOMETHING (reg_info[mcnt]) = 0;
- }
-
- if (regs)
- for (mcnt = 0; mcnt < RE_NREGS; mcnt++)
- regs->start[mcnt] = regs->end[mcnt] = -1;
-
- /* Set up pointers to ends of strings.
- Don't allow the second string to be empty unless both are empty. */
- if (size2 == 0)
- {
- string2 = string1;
- size2 = size1;
- string1 = 0;
- size1 = 0;
- }
- end1 = string1 + size1;
- end2 = string2 + size2;
-
- /* Compute where to stop matching, within the two strings. */
- if (mstop <= size1)
- {
- end_match_1 = string1 + mstop;
- end_match_2 = string2;
- }
- else
- {
- end_match_1 = end1;
- end_match_2 = string2 + mstop - size1;
- }
-
- /* `p' scans through the pattern as `d' scans through the data. `dend'
- is the end of the input string that `d' points within. `d' is
- advanced into the following input string whenever necessary, but
- this happens before fetching; therefore, at the beginning of the
- loop, `d' can be pointing at the end of a string, but it cannot
- equal string2. */
-
- if (size1 != 0 && pos <= size1)
- d = string1 + pos, dend = end_match_1;
- else
- d = string2 + pos - size1, dend = end_match_2;
-
-
- /* This loops over pattern commands. It exits by returning from the
- function if match is complete, or it drops through if match fails
- at this starting point in the input data. */
-
- while (1)
- {
- is_a_jump_n = 0;
- /* End of pattern means we might have succeeded. */
- if (p == pend)
- {
- /* If not end of string, try backtracking. Otherwise done. */
- if (d != end_match_2)
- {
- if (stackp != stackb)
- {
- /* More failure points to try. */
-
- unsigned in_same_string =
- IS_IN_FIRST_STRING (best_regend[0])
- == MATCHING_IN_FIRST_STRING;
-
- /* If exceeds best match so far, save it. */
- if (! best_regs_set
- || (in_same_string && d > best_regend[0])
- || (! in_same_string && ! MATCHING_IN_FIRST_STRING))
- {
- best_regs_set = 1;
- best_regend[0] = d; /* Never use regstart[0]. */
-
- for (mcnt = 1; mcnt < RE_NREGS; mcnt++)
- {
- best_regstart[mcnt] = regstart[mcnt];
- best_regend[mcnt] = regend[mcnt];
- }
- }
- goto fail;
- }
- /* If no failure points, don't restore garbage. */
- else if (best_regs_set)
- {
- restore_best_regs:
- /* Restore best match. */
- d = best_regend[0];
-
- for (mcnt = 0; mcnt < RE_NREGS; mcnt++)
- {
- regstart[mcnt] = best_regstart[mcnt];
- regend[mcnt] = best_regend[mcnt];
- }
- }
- }
-
- /* If caller wants register contents data back, convert it
- to indices. */
- if (regs)
- {
- regs->start[0] = pos;
- if (MATCHING_IN_FIRST_STRING)
- regs->end[0] = d - string1;
- else
- regs->end[0] = d - string2 + size1;
- for (mcnt = 1; mcnt < RE_NREGS; mcnt++)
- {
- if (regend[mcnt] == (unsigned char *) -1)
- {
- regs->start[mcnt] = -1;
- regs->end[mcnt] = -1;
- continue;
- }
- if (IS_IN_FIRST_STRING (regstart[mcnt]))
- regs->start[mcnt] = regstart[mcnt] - string1;
- else
- regs->start[mcnt] = regstart[mcnt] - string2 + size1;
-
- if (IS_IN_FIRST_STRING (regend[mcnt]))
- regs->end[mcnt] = regend[mcnt] - string1;
- else
- regs->end[mcnt] = regend[mcnt] - string2 + size1;
- }
- }
- return d - pos - (MATCHING_IN_FIRST_STRING
- ? string1
- : string2 - size1);
- }
-
- /* Otherwise match next pattern command. */
- #ifdef SWITCH_ENUM_BUG
- switch ((int) ((enum regexpcode) *p++))
- #else
- switch ((enum regexpcode) *p++)
- #endif
- {
-
- /* \( [or `(', as appropriate] is represented by start_memory,
- \) by stop_memory. Both of those commands are followed by
- a register number in the next byte. The text matched
- within the \( and \) is recorded under that number. */
- case start_memory:
- regstart[*p] = d;
- IS_ACTIVE (reg_info[*p]) = 1;
- MATCHED_SOMETHING (reg_info[*p]) = 0;
- p++;
- break;
-
- case stop_memory:
- regend[*p] = d;
- IS_ACTIVE (reg_info[*p]) = 0;
-
- /* If just failed to match something this time around with a sub-
- expression that's in a loop, try to force exit from the loop. */
- if ((! MATCHED_SOMETHING (reg_info[*p])
- || (enum regexpcode) p[-3] == start_memory)
- && (p + 1) != pend)
- {
- register unsigned char *p2 = p + 1;
- mcnt = 0;
- switch (*p2++)
- {
- case jump_n:
- is_a_jump_n = 1;
- case finalize_jump:
- case maybe_finalize_jump:
- case jump:
- case dummy_failure_jump:
- EXTRACT_NUMBER_AND_INCR (mcnt, p2);
- if (is_a_jump_n)
- p2 += 2;
- break;
- }
- p2 += mcnt;
-
- /* If the next operation is a jump backwards in the pattern
- to an on_failure_jump, exit from the loop by forcing a
- failure after pushing on the stack the on_failure_jump's
- jump in the pattern, and d. */
- if (mcnt < 0 && (enum regexpcode) *p2++ == on_failure_jump)
- {
- EXTRACT_NUMBER_AND_INCR (mcnt, p2);
- PUSH_FAILURE_POINT (p2 + mcnt, d);
- goto fail;
- }
- }
- p++;
- break;
-
- /* \<digit> has been turned into a `duplicate' command which is
- followed by the numeric value of <digit> as the register number. */
- case duplicate:
- {
- int regno = *p++; /* Get which register to match against */
- register unsigned char *d2, *dend2;
-
- /* Where in input to try to start matching. */
- d2 = regstart[regno];
-
- /* Where to stop matching; if both the place to start and
- the place to stop matching are in the same string, then
- set to the place to stop, otherwise, for now have to use
- the end of the first string. */
-
- dend2 = ((IS_IN_FIRST_STRING (regstart[regno])
- == IS_IN_FIRST_STRING (regend[regno]))
- ? regend[regno] : end_match_1);
- while (1)
- {
- /* If necessary, advance to next segment in register
- contents. */
- while (d2 == dend2)
- {
- if (dend2 == end_match_2) break;
- if (dend2 == regend[regno]) break;
- d2 = string2, dend2 = regend[regno]; /* end of string1 => advance to string2. */
- }
- /* At end of register contents => success */
- if (d2 == dend2) break;
-
- /* If necessary, advance to next segment in data. */
- PREFETCH;
-
- /* How many characters left in this segment to match. */
- mcnt = dend - d;
-
- /* Want how many consecutive characters we can match in
- one shot, so, if necessary, adjust the count. */
- if (mcnt > dend2 - d2)
- mcnt = dend2 - d2;
-
- /* Compare that many; failure if mismatch, else move
- past them. */
- if (translate
- ? bcmp_translate (d, d2, mcnt, translate)
- : bcmp (d, d2, mcnt))
- goto fail;
- d += mcnt, d2 += mcnt;
- }
- }
- break;
-
- case anychar:
- PREFETCH; /* Fetch a data character. */
- /* Match anything but a newline, maybe even a null. */
- if ((translate ? translate[*d] : *d) == '\n'
- || ((obscure_syntax & RE_DOT_NOT_NULL)
- && (translate ? translate[*d] : *d) == '\000'))
- goto fail;
- SET_REGS_MATCHED;
- d++;
- break;
-
- case charset:
- case charset_not:
- {
- int not = 0; /* Nonzero for charset_not. */
- register int c;
- if (*(p - 1) == (unsigned char) charset_not)
- not = 1;
-
- PREFETCH; /* Fetch a data character. */
-
- if (translate)
- c = translate[*d];
- else
- c = *d;
-
- if (c < *p * BYTEWIDTH
- && p[1 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
- not = !not;
-
- p += 1 + *p;
-
- if (!not) goto fail;
- SET_REGS_MATCHED;
- d++;
- break;
- }
-
- case begline:
- if ((size1 != 0 && d == string1)
- || (size1 == 0 && size2 != 0 && d == string2)
- || (d && d[-1] == '\n')
- || (size1 == 0 && size2 == 0))
- break;
- else
- goto fail;
-
- case endline:
- if (d == end2
- || (d == end1 ? (size2 == 0 || *string2 == '\n') : *d == '\n'))
- break;
- goto fail;
-
- /* `or' constructs are handled by starting each alternative with
- an on_failure_jump that points to the start of the next
- alternative. Each alternative except the last ends with a
- jump to the joining point. (Actually, each jump except for
- the last one really jumps to the following jump, because
- tensioning the jumps is a hassle.) */
-
- /* The start of a stupid repeat has an on_failure_jump that points
- past the end of the repeat text. This makes a failure point so
- that on failure to match a repetition, matching restarts past
- as many repetitions have been found with no way to fail and
- look for another one. */
-
- /* A smart repeat is similar but loops back to the on_failure_jump
- so that each repetition makes another failure point. */
-
- case on_failure_jump:
- on_failure:
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- PUSH_FAILURE_POINT (p + mcnt, d);
- break;
-
- /* The end of a smart repeat has a maybe_finalize_jump back.
- Change it either to a finalize_jump or an ordinary jump. */
- case maybe_finalize_jump:
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- {
- register unsigned char *p2 = p;
- /* Compare what follows with the beginning of the repeat.
- If we can establish that there is nothing that they would
- both match, we can change to finalize_jump. */
- while (p2 + 1 != pend
- && (*p2 == (unsigned char) stop_memory
- || *p2 == (unsigned char) start_memory))
- p2 += 2; /* Skip over reg number. */
- if (p2 == pend)
- p[-3] = (unsigned char) finalize_jump;
- else if (*p2 == (unsigned char) exactn
- || *p2 == (unsigned char) endline)
- {
- register int c = *p2 == (unsigned char) endline ? '\n' : p2[2];
- register unsigned char *p1 = p + mcnt;
- /* p1[0] ... p1[2] are an on_failure_jump.
- Examine what follows that. */
- if (p1[3] == (unsigned char) exactn && p1[5] != c)
- p[-3] = (unsigned char) finalize_jump;
- else if (p1[3] == (unsigned char) charset
- || p1[3] == (unsigned char) charset_not)
- {
- int not = p1[3] == (unsigned char) charset_not;
- if (c < p1[4] * BYTEWIDTH
- && p1[5 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
- not = !not;
- /* `not' is 1 if c would match. */
- /* That means it is not safe to finalize. */
- if (!not)
- p[-3] = (unsigned char) finalize_jump;
- }
- }
- }
- p -= 2; /* Point at relative address again. */
- if (p[-1] != (unsigned char) finalize_jump)
- {
- p[-1] = (unsigned char) jump;
- goto nofinalize;
- }
- /* Note fall through. */
-
- /* The end of a stupid repeat has a finalize_jump back to the
- start, where another failure point will be made which will
- point to after all the repetitions found so far. */
-
- /* Take off failure points put on by matching on_failure_jump
- because didn't fail. Also remove the register information
- put on by the on_failure_jump. */
- case finalize_jump:
- POP_FAILURE_POINT ();
- /* Note fall through. */
-
- /* Jump without taking off any failure points. */
- case jump:
- nofinalize:
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- p += mcnt;
- break;
-
- case dummy_failure_jump:
- /* Normally, the on_failure_jump pushes a failure point, which
- then gets popped at finalize_jump. We will end up at
- finalize_jump, also, and with a pattern of, say, `a+', we
- are skipping over the on_failure_jump, so we have to push
- something meaningless for finalize_jump to pop. */
- PUSH_FAILURE_POINT (0, 0);
- goto nofinalize;
-
-
- /* Have to succeed matching what follows at least n times. Then
- just handle like an on_failure_jump. */
- case succeed_n:
- EXTRACT_NUMBER (mcnt, p + 2);
- /* Originally, this is how many times we HAVE to succeed. */
- if (mcnt)
- {
- mcnt--;
- p += 2;
- STORE_NUMBER_AND_INCR (p, mcnt);
- }
- else if (mcnt == 0)
- {
- p[2] = unused;
- p[3] = unused;
- goto on_failure;
- }
- else
- {
- fprintf (stderr, "regex: the succeed_n's n is not set.\n");
- exit (1);
- }
- break;
-
- case jump_n:
- EXTRACT_NUMBER (mcnt, p + 2);
- /* Originally, this is how many times we CAN jump. */
- if (mcnt)
- {
- mcnt--;
- STORE_NUMBER(p + 2, mcnt);
- goto nofinalize; /* Do the jump without taking off
- any failure points. */
- }
- /* If don't have to jump any more, skip over the rest of command. */
- else
- p += 4;
- break;
-
- case set_number_at:
- {
- register unsigned char *p1;
-
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- p1 = p + mcnt;
- EXTRACT_NUMBER_AND_INCR (mcnt, p);
- STORE_NUMBER (p1, mcnt);
- break;
- }
-
- /* Ignore these. Used to ignore the n of succeed_n's which
- currently have n == 0. */
- case unused:
- break;
-
- case wordbound:
- if (AT_WORD_BOUNDARY)
- break;
- goto fail;
-
- case notwordbound:
- if (AT_WORD_BOUNDARY)
- goto fail;
- break;
-
- case wordbeg:
- if (IS_A_LETTER (d) && (!IS_A_LETTER (d - 1) || AT_STRINGS_BEG))
- break;
- goto fail;
-
- case wordend:
- /* Have to check if AT_STRINGS_BEG before looking at d - 1. */
- if (!AT_STRINGS_BEG && IS_A_LETTER (d - 1)
- && (!IS_A_LETTER (d) || AT_STRINGS_END))
- break;
- goto fail;
-
- #ifdef emacs
- case before_dot:
- if (PTR_CHAR_POS (d) >= point)
- goto fail;
- break;
-
- case at_dot:
- if (PTR_CHAR_POS (d) != point)
- goto fail;
- break;
-
- case after_dot:
- if (PTR_CHAR_POS (d) <= point)
- goto fail;
- break;
-
- case wordchar:
- mcnt = (int) Sword;
- goto matchsyntax;
-
- case syntaxspec:
- mcnt = *p++;
- matchsyntax:
- PREFETCH;
- if (SYNTAX (*d++) != (enum syntaxcode) mcnt) goto fail;
- SET_REGS_MATCHED;
- break;
-
- case notwordchar:
- mcnt = (int) Sword;
- goto matchnotsyntax;
-
- case notsyntaxspec:
- mcnt = *p++;
- matchnotsyntax:
- PREFETCH;
- if (SYNTAX (*d++) == (enum syntaxcode) mcnt) goto fail;
- SET_REGS_MATCHED;
- break;
-
- #else /* not emacs */
-
- case wordchar:
- PREFETCH;
- if (!IS_A_LETTER (d))
- goto fail;
- SET_REGS_MATCHED;
- break;
-
- case notwordchar:
- PREFETCH;
- if (IS_A_LETTER (d))
- goto fail;
- SET_REGS_MATCHED;
- break;
-
- #endif /* not emacs */
-
- case begbuf:
- if (AT_STRINGS_BEG)
- break;
- goto fail;
-
- case endbuf:
- if (AT_STRINGS_END)
- break;
- goto fail;
-
- case exactn:
- /* Match the next few pattern characters exactly.
- mcnt is how many characters to match. */
- mcnt = *p++;
- /* This is written out as an if-else so we don't waste time
- testing `translate' inside the loop. */
- if (translate)
- {
- do
- {
- PREFETCH;
- if (translate[*d++] != *p++) goto fail;
- }
- while (--mcnt);
- }
- else
- {
- do
- {
- PREFETCH;
- if (*d++ != *p++) goto fail;
- }
- while (--mcnt);
- }
- SET_REGS_MATCHED;
- break;
- }
- continue; /* Successfully executed one pattern command; keep going. */
-
- /* Jump here if any matching operation fails. */
- fail:
- if (stackp != stackb)
- /* A restart point is known. Restart there and pop it. */
- {
- short last_used_reg, this_reg;
-
- /* If this failure point is from a dummy_failure_point, just
- skip it. */
- if (!stackp[-2])
- {
- POP_FAILURE_POINT ();
- goto fail;
- }
-
- d = *--stackp;
- p = *--stackp;
- if (d >= string1 && d <= end1)
- dend = end_match_1;
- /* Restore register info. */
- last_used_reg = (short) *--stackp;
-
- /* Make the ones that weren't saved -1 or 0 again. */
- for (this_reg = RE_NREGS - 1; this_reg > last_used_reg; this_reg--)
- {
- regend[this_reg] = (unsigned char *) -1;
- regstart[this_reg] = (unsigned char *) -1;
- IS_ACTIVE (reg_info[this_reg]) = 0;
- MATCHED_SOMETHING (reg_info[this_reg]) = 0;
- }
-
- /* And restore the rest from the stack. */
- for ( ; this_reg > 0; this_reg--)
- {
- reg_info[this_reg] = *(struct register_info *) *--stackp;
- regend[this_reg] = *--stackp;
- regstart[this_reg] = *--stackp;
- }
- }
- else
- break; /* Matching at this starting point really fails. */
- }
-
- if (best_regs_set)
- goto restore_best_regs;
- return -1; /* Failure to match. */
- }
-
-
- static int
- bcmp_translate (s1, s2, len, translate)
- unsigned char *s1, *s2;
- register int len;
- unsigned char *translate;
- {
- register unsigned char *p1 = s1, *p2 = s2;
- while (len)
- {
- if (translate [*p1++] != translate [*p2++]) return 1;
- len--;
- }
- return 0;
- }
-
-
-
- /* Entry points compatible with 4.2 BSD regex library. */
-
- #ifndef emacs
-
- static struct re_pattern_buffer re_comp_buf;
-
- char *
- re_comp (s)
- char *s;
- {
- if (!s)
- {
- if (!re_comp_buf.buffer)
- return "No previous regular expression";
- return 0;
- }
-
- if (!re_comp_buf.buffer)
- {
- if (!(re_comp_buf.buffer = (char *) malloc (200)))
- return "Memory exhausted";
- re_comp_buf.allocated = 200;
- if (!(re_comp_buf.fastmap = (char *) malloc (1 << BYTEWIDTH)))
- return "Memory exhausted";
- }
- return re_compile_pattern (s, strlen (s), &re_comp_buf);
- }
-
- int
- re_exec (s)
- char *s;
- {
- int len = strlen (s);
- return 0 <= re_search (&re_comp_buf, s, len, 0, len,
- (struct re_registers *) 0);
- }
- #endif /* not emacs */
-
-
-
- #ifdef test
-
- #include <stdio.h>
-
- /* Indexed by a character, gives the upper case equivalent of the
- character. */
-
- char upcase[0400] =
- { 000, 001, 002, 003, 004, 005, 006, 007,
- 010, 011, 012, 013, 014, 015, 016, 017,
- 020, 021, 022, 023, 024, 025, 026, 027,
- 030, 031, 032, 033, 034, 035, 036, 037,
- 040, 041, 042, 043, 044, 045, 046, 047,
- 050, 051, 052, 053, 054, 055, 056, 057,
- 060, 061, 062, 063, 064, 065, 066, 067,
- 070, 071, 072, 073, 074, 075, 076, 077,
- 0100, 0101, 0102, 0103, 0104, 0105, 0106, 0107,
- 0110, 0111, 0112, 0113, 0114, 0115, 0116, 0117,
- 0120, 0121, 0122, 0123, 0124, 0125, 0126, 0127,
- 0130, 0131, 0132, 0133, 0134, 0135, 0136, 0137,
- 0140, 0101, 0102, 0103, 0104, 0105, 0106, 0107,
- 0110, 0111, 0112, 0113, 0114, 0115, 0116, 0117,
- 0120, 0121, 0122, 0123, 0124, 0125, 0126, 0127,
- 0130, 0131, 0132, 0173, 0174, 0175, 0176, 0177,
- 0200, 0201, 0202, 0203, 0204, 0205, 0206, 0207,
- 0210, 0211, 0212, 0213, 0214, 0215, 0216, 0217,
- 0220, 0221, 0222, 0223, 0224, 0225, 0226, 0227,
- 0230, 0231, 0232, 0233, 0234, 0235, 0236, 0237,
- 0240, 0241, 0242, 0243, 0244, 0245, 0246, 0247,
- 0250, 0251, 0252, 0253, 0254, 0255, 0256, 0257,
- 0260, 0261, 0262, 0263, 0264, 0265, 0266, 0267,
- 0270, 0271, 0272, 0273, 0274, 0275, 0276, 0277,
- 0300, 0301, 0302, 0303, 0304, 0305, 0306, 0307,
- 0310, 0311, 0312, 0313, 0314, 0315, 0316, 0317,
- 0320, 0321, 0322, 0323, 0324, 0325, 0326, 0327,
- 0330, 0331, 0332, 0333, 0334, 0335, 0336, 0337,
- 0340, 0341, 0342, 0343, 0344, 0345, 0346, 0347,
- 0350, 0351, 0352, 0353, 0354, 0355, 0356, 0357,
- 0360, 0361, 0362, 0363, 0364, 0365, 0366, 0367,
- 0370, 0371, 0372, 0373, 0374, 0375, 0376, 0377
- };
-
- #ifdef canned
-
- #include "tests.h"
-
- typedef enum { extended_test, basic_test } test_type;
-
- /* Use this to run the tests we've thought of. */
-
- void
- main ()
- {
- test_type t = extended_test;
-
- if (t == basic_test)
- {
- printf ("Running basic tests:\n\n");
- test_posix_basic ();
- }
- else if (t == extended_test)
- {
- printf ("Running extended tests:\n\n");
- test_posix_extended ();
- }
- }
-
- #else /* not canned */
-
- /* Use this to run interactive tests. */
-
- void
- main (argc, argv)
- int argc;
- char **argv;
- {
- char pat[80];
- struct re_pattern_buffer buf;
- int i;
- char c;
- char fastmap[(1 << BYTEWIDTH)];
-
- /* Allow a command argument to specify the style of syntax. */
- if (argc > 1)
- obscure_syntax = atoi (argv[1]);
-
- buf.allocated = 40;
- buf.buffer = (char *) malloc (buf.allocated);
- buf.fastmap = fastmap;
- buf.translate = upcase;
-
- while (1)
- {
- gets (pat);
-
- if (*pat)
- {
- re_compile_pattern (pat, strlen(pat), &buf);
-
- for (i = 0; i < buf.used; i++)
- printchar (buf.buffer[i]);
-
- putchar ('\n');
-
- printf ("%d allocated, %d used.\n", buf.allocated, buf.used);
-
- re_compile_fastmap (&buf);
- printf ("Allowed by fastmap: ");
- for (i = 0; i < (1 << BYTEWIDTH); i++)
- if (fastmap[i]) printchar (i);
- putchar ('\n');
- }
-
- gets (pat); /* Now read the string to match against */
-
- i = re_match (&buf, pat, strlen (pat), 0, 0);
- printf ("Match value %d.\n", i);
- }
- }
-
- #endif
-
-
- #ifdef NOTDEF
- print_buf (bufp)
- struct re_pattern_buffer *bufp;
- {
- int i;
-
- printf ("buf is :\n----------------\n");
- for (i = 0; i < bufp->used; i++)
- printchar (bufp->buffer[i]);
-
- printf ("\n%d allocated, %d used.\n", bufp->allocated, bufp->used);
-
- printf ("Allowed by fastmap: ");
- for (i = 0; i < (1 << BYTEWIDTH); i++)
- if (bufp->fastmap[i])
- printchar (i);
- printf ("\nAllowed by translate: ");
- if (bufp->translate)
- for (i = 0; i < (1 << BYTEWIDTH); i++)
- if (bufp->translate[i])
- printchar (i);
- printf ("\nfastmap is%s accurate\n", bufp->fastmap_accurate ? "" : "n't");
- printf ("can %s be null\n----------", bufp->can_be_null ? "" : "not");
- }
- #endif /* NOTDEF */
-
- printchar (c)
- char c;
- {
- if (c < 040 || c >= 0177)
- {
- putchar ('\\');
- putchar (((c >> 6) & 3) + '0');
- putchar (((c >> 3) & 7) + '0');
- putchar ((c & 7) + '0');
- }
- else
- putchar (c);
- }
-
- error (string)
- char *string;
- {
- puts (string);
- exit (1);
- }
- #endif /* test */
-