Shapefile Parsing Issue

This issue still exists in 6.x-1.0-alpha2. I looked into applying the wkb changes to the shp2sql file, but my attempts all ended in failure; admittedly I don't know a ton about how to properly use unpack and byte_order.

A coworker and I (scottschulthess, at OHO Interactive) were able to update shp2sql.inc and successfully import both polylines and polygons. The bulk of the clues we found were from this article: http://www.easywms.com/easywms/?q=zh-hant/node/78
Here are the things I believe we changed:

1. _shp_get_linestring was not pulling the numparts (4 bytes) out of the file, so the bytes were off when loading a polyline into the database. This was causing the geo_type to come up as unsupported. (I'm not sure if a polyline/linestring can ever have more than one numpart, so right now we're assuming its only one)
2. The unpack parameter 'i' was changed to 'V'. 'i' is machine-dependent, but 'V' will always read in little endian byte-order.
3. in _shp_get_linestring, the wkt is built but never added to the $data array which is returned. This results in a null geo object.
4. There seems to be some issue with the way the unpack method is used in a couple places. in _shp_get_linestring, it pulls 4 8-byte elements from the file (as one 32 byte segment). When we changed this to 4 separate 8-byte reads, we were able to successfully import. I don't really understand the difference between what we wrote and what was already there, so I've pasted it here in case anyone has any ideas:

Original: (broken)

  $data = array(
    'bbox'  => unpack('dmin_x/dmin_y/dmax_x/dmax_y', zip_entry_read($fp, 32)),
    'point_count' => current(unpack('i', zip_entry_read($fp, 4))),
  );

Original (patched): (also broken)

  $data = array(
    'bbox'  => unpack('dmin_x/dmin_y/dmax_x/dmax_y', zip_entry_read($fp, 32)),
    'part_count' => current(unpack('V', zip_entry_read($fp, 4))),
    'point_count' => current(unpack('V', zip_entry_read($fp, 4))),
  );

New code: (working)

  $min_x = loadData('d', zip_entry_read($fp, 8));
  $min_y = loadData('d', zip_entry_read($fp, 8));
  $max_x = loadData('d', zip_entry_read($fp, 8));
  $max_y = loadData('d', zip_entry_read($fp, 8));
  $numparts = loadData('V', zip_entry_read($fp, 4));
  $numpoints= loadData('V', zip_entry_read($fp, 4));

loadData func:

function loadData($format, $data) {
  if (!$data) return $data;
  $tmp = unpack($format, $data);
  return current($tmp);
}

I've uploaded our modified shp2sql.inc and would be happy to commit it to CVS if the maintainer approves.

I've also just realized that 'd' is also a machine-dependent format specifier. If anyone is on a different architecture (powerpc most likely) and having issues, I'd try using 'V' in place of 'd'.

Changing status to needs review