Both solutions seems to work, however there are caveats for each one given I wasn't very specific on the requirements.
new data set:

Code:

10a textile
b10b wtf
b20b omg
b20b woot
20b teasdf ha
20b tesdf2 he
30c woot1
30c woot2

radoulov solutions works, though it assumes integers. I like the simplicity so I've tried modifying this code to work for non-integer indices, and it seems to lose the row order significantly(row to column transposed order is OK though):

Code:

END {
  for (i in _) 
    printf "%s\n", _[i]
  }
{
  _[$1] = _[$1] ? _[$1] FS $0 : $0
}

This implies a problem with the string as array indices, but ivhb's solution uses the same feature.

Code:

{
  if (cnt[$1] ++ == 0)
  {
    idx[j++] = $1
     rst[$1] = $0
  } else {
     rst[$1] = rst[$1] FS $0 
  }
}
END {
  for (i = 0; i < j; i ++)
    print rst[idx[i]]
}

This works OK as modified.

Final script FTW:

Code:

{
  if (cnt[$1] ++ == 0)
  {
    idx[j++] = $1
     rst[$1] = $0
  } else {
    vindex = $1
     sub(/^[[:alnum:]_]+ \y/, "") 
     rst[vindex] = rst[vindex] FS $0 
  }
}
END {
  for (i = 0; i < j; i ++)
    print rst[idx[i]]
    #print idx[i], rst[idx[i]]
}

The only problem remains is that the space is assumed as the FS in the regexp, is there a way to escape the FS in a regexp?