Re: [FYI] Forward movement on 9404

[Nathan Myers <ncm-nospam at cantrip dot org>]

On Fri, Feb 21, 2003 at 06:07:26PM -0600, Benjamin Kosnik wrote:
> A minor nit:
> 
> +	      if (_M_out_buf_size())
>  		__ret = this->sputc(traits_type::to_char_type(__c));
>  	      else if (__len <= _M_string.max_size())
>  		{
>  		  // Force-allocate, re-sync.
>  		  _M_string = this->str();
>  		  _M_string.reserve(__len);
> -		  this->_M_buf_size = __len;
> +		  this->_M_buf_size = _M_string.capacity();
>  		  _M_really_sync(this->_M_in_cur - this->_M_in_beg, 
>  				 this->_M_out_cur - this->_M_out_beg);
>  		  __ret = this->sputc(traits_type::to_char_type(__c));
> 
> is there a way to coalesce these
> 
> __ret = this->sputc(traits_type::to_char_type(__c));
> 
> lines?

Yes.  First, clean up the __testout tests:

  if (__builtin_expect(!__testout, false))
    return traits_type::eof();
  if (__builtin_expect(__testeof, false))
    return traits_type::not_eof();

  __size_type __len = std::max(this->_M_buf_size, this->_M_buf_size_opt);
  __len *= 2;

  if (__builtin_expect(!__testwrite && __len > _M_string.max_size(), false))
    return traits_type::eof();
  if (!__testwrite)
    {
      // Force-allocate, re-sync
      ...
    }
  return this->sputc(traits_type::to_char_type(__c));

But can we let the basic_string<> figure out a new length for itself,
and allocate and copy, by calling capacity(__len)?

Nathan Myers
ncm-nospam at cantrip dot org