label.cc

早期freebsd实现

}

void at_expr::evaluate(int tentative, const reference &ref,
		       string &result, substring_position &)
{
  if (tentative)
    ref.canonicalize_authors(result);
  else {
    const char *end, *start = ref.get_authors(&end);
    if (start)
      result.append(start, end - start);
  }
}

void format_expr::evaluate(int tentative, const reference &ref,
			   string &result, substring_position &)
{
  if (tentative)
    return;
  const label_info *lp = ref.get_label_ptr();
  int num = lp == 0 ? ref.get_number() : lp->count;
  if (type != '0')
    result += format_serial(type, num + 1);
  else {
    const char *ptr = itoa(num + first_number);
    int pad = width - strlen(ptr);
    while (--pad >= 0)
      result += '0';
    result += ptr;
  }
}

static const char *format_serial(char c, int n)
{
  assert(n > 0);
  static char buf[128]; // more than enough.
  switch (c) {
  case 'i':
  case 'I':
    {
      char *p = buf;
      // troff uses z and w to represent 10000 and 5000 in Roman
      // numerals; I can find no historical basis for this usage
      const char *s = c == 'i' ? "zwmdclxvi" : "ZWMDCLXVI";
      if (n >= 40000)
	return itoa(n);
      while (n >= 10000) {
	*p++ = s[0];
	n -= 10000;
      }
      for (int i = 1000; i > 0; i /= 10, s += 2) {
	int m = n/i;
	n -= m*i;
	switch (m) {
	case 3:
	  *p++ = s[2];
	  /* falls through */
	case 2:
	  *p++ = s[2];
	  /* falls through */
	case 1:
	  *p++ = s[2];
	  break;
	case 4:
	  *p++ = s[2];
	  *p++ = s[1];
	  break;
	case 8:
	  *p++ = s[1];
	  *p++ = s[2];
	  *p++ = s[2];
	  *p++ = s[2];
	  break;
	case 7:
	  *p++ = s[1];
	  *p++ = s[2];
	  *p++ = s[2];
	  break;
	case 6:
	  *p++ = s[1];
	  *p++ = s[2];
	  break;
	case 5:
	  *p++ = s[1];
	  break;
	case 9:
	  *p++ = s[2];
	  *p++ = s[0];
	}
      }
      *p = 0;
      break;
    }
  case 'a':
  case 'A':
    {
      char *p = buf;
      // this is derived from troff/reg.c
      while (n > 0) {
	int d = n % 26;
	if (d == 0)
	  d = 26;
	n -= d;
	n /= 26;
	*p++ = c + d - 1;	// ASCII dependent
      }
      *p-- = 0;
      // Reverse it.
      char *q = buf;
      while (q < p) {
	char temp = *q;
	*q = *p;
	*p = temp;
	--p;
	++q;
      }
      break;
    }
  default:
    assert(0);
  }
  return buf;
}

void field_expr::evaluate(int, const reference &ref,
			  string &result, substring_position &)
{
  const char *end;
  const char *start = ref.get_field(name, &end);
  if (start) {
    start = nth_field(number, start, &end);
    if (start)
      result.append(start, end - start);
  }
}

void literal_expr::evaluate(int, const reference &,
			    string &result, substring_position &)
{
  result += s;
}

analyzed_expr::analyzed_expr(expression *e)
: unary_expr(e), flags(e ? e->analyze() : 0)
{
}

void analyzed_expr::evaluate(int tentative, const reference &ref,
			     string &result, substring_position &pos)
{
  if (expr)
    expr->evaluate(tentative, ref, result, pos);
}

void star_expr::evaluate(int tentative, const reference &ref,
			 string &result, substring_position &pos)
{
  const label_info *lp = ref.get_label_ptr();
  if (!tentative
      && (lp == 0 || lp->total > 1)
      && expr)
    expr->evaluate(tentative, ref, result, pos);
}

void separator_expr::evaluate(int tentative, const reference &ref,
			      string &result, substring_position &pos)
{
  int start_length = result.length();
  int is_first = pos.start < 0;
  if (expr)
    expr->evaluate(tentative, ref, result, pos);
  if (is_first) {
    pos.start = start_length;
    pos.length = result.length() - start_length;
  }
}

void map_expr::evaluate(int tentative, const reference &ref,
			string &result, substring_position &)
{
  if (expr) {
    string temp;
    substring_position temp_pos;
    expr->evaluate(tentative, ref, temp, temp_pos);
    (*func)(temp.contents(), temp.contents() + temp.length(), result);
  }
}

void extractor_expr::evaluate(int tentative, const reference &ref,
			      string &result, substring_position &)
{
  if (expr) {
    string temp;
    substring_position temp_pos;
    expr->evaluate(tentative, ref, temp, temp_pos);
    const char *end, *start = (*func)(temp.contents(),
				      temp.contents() + temp.length(),
				      &end);
    switch (part) {
    case BEFORE:
      if (start)
	result.append(temp.contents(), start - temp.contents());
      else
	result += temp;
      break;
    case MATCH:
      if (start)
	result.append(start, end - start);
      break;
    case AFTER:
      if (start)
	result.append(end, temp.contents() + temp.length() - end);
      break;
    default:
      assert(0);
    }
  }
}

static void first_part(int len, const char *ptr, const char *end,
			  string &result)
{
  for (;;) {
    const char *token_start = ptr;
    if (!get_token(&ptr, end))
      break;
    const token_info *ti = lookup_token(token_start, ptr);
    int counts = ti->sortify_non_empty(token_start, ptr);
    if (counts && --len < 0)
      break;
    if (counts || ti->is_accent())
      result.append(token_start, ptr - token_start);
  }
}

static void last_part(int len, const char *ptr, const char *end,
		      string &result)
{
  const char *start = ptr;
  int count = 0;
  for (;;) {
    const char *token_start = ptr;
    if (!get_token(&ptr, end))
      break;
    const token_info *ti = lookup_token(token_start, ptr);
    if (ti->sortify_non_empty(token_start, ptr))
      count++;
  }
  ptr = start;
  int skip = count - len;
  if (skip > 0) {
    for (;;) {
      const char *token_start = ptr;
      if (!get_token(&ptr, end))
	assert(0);
      const token_info *ti = lookup_token(token_start, ptr);
      if (ti->sortify_non_empty(token_start, ptr) && --skip < 0) {
	ptr = token_start;
	break;
      }
    }
  }
  first_part(len, ptr, end, result);
}

void truncate_expr::evaluate(int tentative, const reference &ref,
			     string &result, substring_position &)
{
  if (expr) {
    string temp;
    substring_position temp_pos;
    expr->evaluate(tentative, ref, temp, temp_pos);
    const char *start = temp.contents();
    const char *end = start + temp.length();
    if (n > 0)
      first_part(n, start, end, result);
    else if (n < 0)
      last_part(-n, start, end, result);
  }
}

void alternative_expr::evaluate(int tentative, const reference &ref,
				string &result, substring_position &pos)
{
  int start_length = result.length();
  if (expr1)
    expr1->evaluate(tentative, ref, result, pos);
  if (result.length() == start_length && expr2)
    expr2->evaluate(tentative, ref, result, pos);
}

void list_expr::evaluate(int tentative, const reference &ref,
			 string &result, substring_position &pos)
{
  if (expr1)
    expr1->evaluate(tentative, ref, result, pos);
  if (expr2)
    expr2->evaluate(tentative, ref, result, pos);
}

void substitute_expr::evaluate(int tentative, const reference &ref,
			       string &result, substring_position &pos)
{
  int start_length = result.length();
  if (expr1)
    expr1->evaluate(tentative, ref, result, pos);
  if (result.length() > start_length && result[result.length() - 1] == '-') {
    // ought to see if pos covers the -
    result.set_length(result.length() - 1);
    if (expr2)
      expr2->evaluate(tentative, ref, result, pos);
  }
}

void conditional_expr::evaluate(int tentative, const reference &ref,
				string &result, substring_position &pos)
{
  string temp;
  substring_position temp_pos;
  if (expr1)
    expr1->evaluate(tentative, ref, temp, temp_pos);
  if (temp.length() > 0) {
    if (expr2)
      expr2->evaluate(tentative, ref, result, pos);
  }
  else {
    if (expr3)
      expr3->evaluate(tentative, ref, result, pos);
  }
}

void reference::pre_compute_label()
{
  if (parsed_label != 0
      && (parsed_label->analyze() & expression::CONTAINS_VARIABLE)) {
    label.clear();
    substring_position temp_pos;
    parsed_label->evaluate(1, *this, label, temp_pos);
    label_ptr = lookup_label(label);
  }
}

void reference::compute_label()
{
  label.clear();
  if (parsed_label)
    parsed_label->evaluate(0, *this, label, separator_pos);
  if (short_label_flag && parsed_short_label)
    parsed_short_label->evaluate(0, *this, short_label, short_separator_pos);
  if (date_as_label) {
    string new_date;
    if (parsed_date_label) {
      substring_position temp_pos;
      parsed_date_label->evaluate(0, *this, new_date, temp_pos);
    }
    set_date(new_date);
  }
  if (label_ptr)
    label_ptr->count += 1;
}

void reference::immediate_compute_label()
{
  if (label_ptr)
    label_ptr->total = 2;	// force use of disambiguator
  compute_label();
}

int reference::merge_labels(reference **v, int n, label_type type,
			    string &result)
{
  if (abbreviate_label_ranges)
    return merge_labels_by_number(v, n, type, result);
  else
    return merge_labels_by_parts(v, n, type, result);
}

int reference::merge_labels_by_number(reference **v, int n, label_type type,
				      string &result)
{
  if (n <= 1)
    return 0;
  int num = get_number();
  // Only merge three or more labels.
  if (v[0]->get_number() != num + 1
      || v[1]->get_number() != num + 2)
    return 0;
  for (int i = 2; i < n; i++)
    if (v[i]->get_number() != num + i + 1)
      break;
  result = get_label(type);
  result += label_range_indicator;
  result += v[i - 1]->get_label(type);
  return i;
}

const substring_position &reference::get_separator_pos(label_type type) const
{
  if (type == SHORT_LABEL && short_label_flag)
    return short_separator_pos;
  else
    return separator_pos;
}

const string &reference::get_label(label_type type) const
{
  if (type == SHORT_LABEL && short_label_flag)
    return short_label; 
  else
    return label;
}

int reference::merge_labels_by_parts(reference **v, int n, label_type type,
				     string &result)
{
  if (n <= 0)
    return 0;
  const string &lb = get_label(type);
  const substring_position &sp = get_separator_pos(type);
  if (sp.start < 0
      || sp.start != v[0]->get_separator_pos(type).start 
      || memcmp(lb.contents(), v[0]->get_label(type).contents(),
		sp.start) != 0)
    return 0;
  result = lb;
  int i = 0;
  do {
    result += separate_label_second_parts;
    const substring_position &s = v[i]->get_separator_pos(type);
    int sep_end_pos = s.start + s.length;
    result.append(v[i]->get_label(type).contents() + sep_end_pos,
		  v[i]->get_label(type).length() - sep_end_pos);
  } while (++i < n
	   && sp.start == v[i]->get_separator_pos(type).start
	   && memcmp(lb.contents(), v[i]->get_label(type).contents(),
		     sp.start) == 0);
  return i;
}

string label_pool;

label_info::label_info(const string &s)
: count(0), total(1), length(s.length()), start(label_pool.length())
{
  label_pool += s;
}

static label_info **label_table = 0;
static int label_table_size = 0;
static int label_table_used = 0;

label_info *lookup_label(const string &label)
{
  if (label_table == 0) {
    label_table = new label_info *[17];
    label_table_size = 17;
    for (int i = 0; i < 17; i++)
      label_table[i] = 0;
  }
  unsigned h = hash_string(label.contents(), label.length()) % label_table_size;
  for (label_info **ptr = label_table + h;
       *ptr != 0;
       (ptr == label_table)
       ? (ptr = label_table + label_table_size - 1)
       : ptr--)
    if ((*ptr)->length == label.length()
	&& memcmp(label_pool.contents() + (*ptr)->start, label.contents(),
		  label.length()) == 0) {
      (*ptr)->total += 1;
      return *ptr;
    }
  label_info *result = *ptr = new label_info(label);
  if (++label_table_used * 2 > label_table_size) {
    // Rehash the table.
    label_info **old_table = label_table;
    int old_size = label_table_size;
    label_table_size = next_size(label_table_size);
    label_table = new label_info *[label_table_size];
    int i;
    for (i = 0; i < label_table_size; i++)
      label_table[i] = 0;
    for (i = 0; i < old_size; i++)
      if (old_table[i]) {
	unsigned h = hash_string(label_pool.contents() + old_table[i]->start,
				 old_table[i]->length);
	for (label_info **p = label_table + (h % label_table_size);
	     *p != 0;
	     (p == label_table)
	     ? (p = label_table + label_table_size - 1)
	     : --p)
	    ;
	*p = old_table[i];
	}
    a_delete old_table;
  }
  return result;
}

void clear_labels()
{
  for (int i = 0; i < label_table_size; i++) {
    delete label_table[i];
    label_table[i] = 0;
  }
  label_table_used = 0;
  label_pool.clear();
}

static void consider_authors(reference **start, reference **end, int i);

void compute_labels(reference **v, int n)
{
  if (parsed_label
      && (parsed_label->analyze() & expression::CONTAINS_AT)
      && sort_fields.length() >= 2
      && sort_fields[0] == 'A'
      && sort_fields[1] == '+')
    consider_authors(v, v + n, 0);
  for (int i = 0; i < n; i++)
    v[i]->compute_label();
}


/* A reference with a list of authors <A0,A1,...,AN> _needs_ author i
where 0 <= i <= N if there exists a reference with a list of authors
<B0,B1,...,BM> such that <A0,A1,...,AN> != <B0,B1,...,BM> and M >= i
and Aj = Bj for 0 <= j < i. In this case if we can't say ``A0,
A1,...,A(i-1) et al'' because this would match both <A0,A1,...,AN> and
<B0,B1,...,BM>.  If a reference needs author i we only have to call
need_author(j) for some j >= i such that the reference also needs
author j. */