1#@!#!123s

: / usr / share / systemtap / tapset /
Filename : regex.stp
back
// Regular expression subexpression tapset
// Copyright (C) 2017 Serhei Makarov
// Copyright (C) 2013 Red Hat, Inc.
//
// This file is part of systemtap, and is free software.  You can
// redistribute it and/or modify it under the terms of the GNU General
// Public License (GPL); either version 2, or (at your option) any
// later version.

%{
#define STAP_NEED_CONTEXT_SUBEXPRESSION 1
%}

/**
 * sfunction matched_str - Return the last matched string.
 *
 * Description: returns the string matched by the last successful
 * use of the =~ regexp matching operator. Returns an error if the
 * last use of =~ led to a failed match.
 */
function matched_str:string() { return matched(0) }

/**
 * sfunction matched - Return a given matched subexpression.
 *
 * @n: index to the subexpression to return. 0 corresponds to the
 * entire regular expression.
 *
 * Description: returns the content of the n'th subexpression of the
 * last successful use of the =~ regex matching operator. Returns an
 * empty string if the n'th subexpression was not matched (e.g. due to
 * alternation). Throws an error if the last use of =~ was a failed
 * match, or if fewer than n subexpressions are present in the
 * original regexp.
 */
function matched:string(n:long)
%{ /* pure */ /* unprivileged */ /* pragma:tagged_dfa */
  int start_ix, end_ix; // indices into tag buffer
  int start, end, length; // actual coordinate values

  if (!CONTEXT->last_match.result) {
    snprintf(CONTEXT->error_buffer, sizeof(CONTEXT->error_buffer),
             "Attempted to get subexpression %lld from failed match", (long long) STAP_ARG_n);
    CONTEXT->last_error = CONTEXT->error_buffer;
  }

  start_ix = 2 * STAP_ARG_n, end_ix = start_ix + 1;

  if (end_ix >= CONTEXT->last_match.num_final_tags) {
    snprintf(CONTEXT->error_buffer, sizeof(CONTEXT->error_buffer),
             "Attempted to get nonexistent subexpression %lld", (long long) STAP_ARG_n);
    CONTEXT->last_error = CONTEXT->error_buffer;
  }

  start = CONTEXT->last_match.tag_vals[start_ix];
  end = CONTEXT->last_match.tag_vals[end_ix];
  // _stp_printf ("**DEBUG** Extracted subexpression #%lld:(%d,%d) from %d to %d\n", STAP_ARG_n, start_ix, end_ix, start, end);

  if (start < 0 || end < 0) {
    // If indices are negative, the group was not matched. Return empty string:
    start = end = 0;
    // snprintf(CONTEXT->error_buffer, sizeof(CONTEXT->error_buffer),
    //          "Unknown coordinates for subexpression %lld", STAP_ARG_n);
    // CONTEXT->last_error = CONTEXT->error_buffer;
  }

  if (start > end) {
    // This should not happen.
    snprintf(CONTEXT->error_buffer, sizeof(CONTEXT->error_buffer),
             "BUG: inverted coordinates for subexpression %lld", (long long) STAP_ARG_n);
    CONTEXT->last_error = CONTEXT->error_buffer;
  }

  length = end - start;

  // TODOXXX assert (start <= strlen(matched_str)) ??
  // XXX: Must add 1 to length to account for NUL byte in strlcpy().
  strlcpy(STAP_RETVALUE, CONTEXT->last_match.matched_str + start, length + 1);
%}

/**
 * sfunction ngroups - Number of subexpressions in the last match.
 *
 * Description: returns the number of subexpressions from the
 * last successful use of the =~ regex matching operator.
 *
 * Note that this number includes subexpressions which are present in
 * the regex but did not match any string; for example, given the
 * regex "a|(b)", the subexpressions will count the group for (b)
 * regardless of whether it matched a string or not. Throws an error
 * if the last use of =~ was a failed match.
 */
function ngroups:long()
%{ /* pure */ /* unprivileged */ /* pragma:tagged_dfa */
  if (!CONTEXT->last_match.result) {
    snprintf(CONTEXT->error_buffer, sizeof(CONTEXT->error_buffer),
             "Attempted to get subexpression count from failed match");
    CONTEXT->last_error = CONTEXT->error_buffer;
  }

  STAP_RETVALUE = CONTEXT->last_match.num_final_tags / 2;
%}

// XXX: perhaps implement matched_start, matched_end to get indices?
// XXX: some kind of find-replace functionality?
// XXX: some kind of splitting / multiple-match functionality?