public Page(int fieldToSortBy, string[] records)
{
int i;
for (i = 0; i < records.Length && records[i] != null; i++)
{
string[] fieldsToParse = records[i].Split(',');
if (_records == null)
{
_numberOfFields = fieldsToParse.Length;
_records = new Record[records.Length];
}
int[] fields = new int[_numberOfFields];
for (int j = 0; j < _numberOfFields; j++)
{
fields[j] = int.Parse(fieldsToParse[j]);
}
_records[i] = new Record(fieldToSortBy ,fields);
}
}
public string Sort(string filename, int fieldIndexToSortBy, int fieldCount)
{
FirstIteration(filename, fieldIndexToSortBy);
while (_iterationToSortedruns[_iteration - 1].Count > 1)
{
Console.WriteLine("\nCommencing iteration {0} \n", _iteration);
int sortedrunIndex = 0;
int sortedrunsUsed = 0;
_pool = new Page[_numberOfBufferPages];
// load the first page of each of the B-1 sortedruns from the last iteration
// and merge sort them using the last page in the buffer pool
for (int i = 0; i < _iterationToSortedruns[_iteration - 1].Count; i++)
{
if (i == _numberOfBufferPages - 1) // Use only B-1 sortedruns
{
sortedrunsUsed = i;
break;
}
_pool[i] = new Page(fieldIndexToSortBy, ReadPage(_iterationToSortedruns[_iteration - 1][i].GetReader()));
}
// point to current index in each sortedrun and also the buffer page
int[] pointers = new int[_numberOfBufferPages];
// reset the buffer page to contains 0 records
_pool[_numberOfBufferPages - 1] = new Page(fieldCount, _pageSizeInRecords);
// initialize the current iteration's pages
int producedSortedruns = 0;
_iterationToSortedruns[_iteration] = new List<SortedRun>();
string newSortedrunFilename = Path.GetTempFileName();
var sortedRun = new SortedRun();
sortedRun.Filename = newSortedrunFilename;
sortedRun.GetWriter();
_iterationToSortedruns[_iteration].Add(sortedRun);
int pageCount = 0;
bool finishedMergingSet = false;
// while there are still sortedruns from previous iteration
while (!finishedMergingSet)
{
finishedMergingSet = true;
for (int i = 0; i < _numberOfBufferPages - 1; i++)
{
if (_pool[i] != null && _pool[i].Records != null && _pool[i].Records[pointers[i]] != null)
{
finishedMergingSet = false;
break;
}
}
if (finishedMergingSet)
{
// close the latest sortedrun written
_iterationToSortedruns[_iteration][producedSortedruns++].Writer.Close();
// load the first page from B-1 sortedruns of the previous iteration
// continuing from the next sortedrun which was not read
sortedrunIndex = sortedrunsUsed;
for (int i = sortedrunsUsed; i < _iterationToSortedruns[_iteration - 1].Count; i++)
{
if (i - sortedrunIndex == (_numberOfBufferPages - 1)) // Use up to B-1 sortedruns
{
break;
}
_pool[i - sortedrunIndex] = new Page(fieldIndexToSortBy, ReadPage(_iterationToSortedruns[_iteration - 1][i].GetReader()));
sortedrunsUsed++;
}
// no more sorted runs?
if (sortedrunIndex == sortedrunsUsed)
{
continue;
}
foreach (var item in _pool.Take(_numberOfBufferPages - 1))
{
if (item != null && item.Records != null)
{
finishedMergingSet = false;
}
}
if (finishedMergingSet)
continue;
// there're more sorted runs, prepare a new merged sortedrun
String tempFilename = Path.GetTempFileName();
_iterationToSortedruns[_iteration].Add(new SortedRun() { Filename = tempFilename });
pageCount = 0;
}
Record smallest = null;
// while the buffer page doesn't exceed
while (pointers[_numberOfBufferPages-1] < _pageSizeInRecords)
{
// initialize the lowest records
smallest = null;
int firstPageThatIsNotEmpty = 0;
// find the first page that is not empty
for (int i = 0; i < _numberOfBufferPages; i++)
{
if (_pool[i] != null && _pool[i].Records != null && _pool[i].Records[pointers[i]] != null)
{
// choose the first non-null record as the smallest
smallest = new Record(_pool[i].Records[pointers[i]]);
firstPageThatIsNotEmpty = i;
break;
}
}
// ran out of values?
if (smallest == null)
break;
int pageIndexContainingLowestKey = firstPageThatIsNotEmpty;
// search for the smallest key in any of the current pages
for (int i = firstPageThatIsNotEmpty + 1; i < _numberOfBufferPages; i++)
{
if (_pool[i] != null && _pool[i].Records != null
&& _pool[i].Records[pointers[i]] != null
&& _pool[i].Records[pointers[i]].Fields[fieldIndexToSortBy] < smallest[fieldIndexToSortBy])
{
smallest = _pool[i].Records[pointers[i]];
pageIndexContainingLowestKey = i;
}
}
// insert the smallest record into the buffer page and increase the page's pointers.
_pool[_numberOfBufferPages - 1].Records[pointers[_numberOfBufferPages-1]] = smallest;
pointers[_numberOfBufferPages-1]++;
pointers[pageIndexContainingLowestKey]++;
if (pointers[pageIndexContainingLowestKey] == _pageSizeInRecords)
{
// get the next page of the sortedrun, and start the count over.
_pool[pageIndexContainingLowestKey] = new Page(fieldIndexToSortBy,
ReadPage(_iterationToSortedruns[_iteration - 1][sortedrunIndex + pageIndexContainingLowestKey].GetReader()));
pointers[pageIndexContainingLowestKey] = 0;
}
}
// completed another page, write it to the current sortedrun in the current iteration
_pool[_numberOfBufferPages - 1].WritePage(_iterationToSortedruns[_iteration][producedSortedruns].GetWriter());
pageCount++;
Console.WriteLine("Wrote page {0} of sorted run #{1}", pageCount, producedSortedruns);
// reset the buffer page to contains 0 records
_pool[_numberOfBufferPages - 1] = new Page(fieldCount, _pageSizeInRecords);
pointers[_numberOfBufferPages - 1] = 0;
}
_iteration++;
}
return _iterationToSortedruns.Last().Value.Last().Filename;
}
public Record(Record record)
{
FieldToSortBy = record.FieldToSortBy;
_fields = record.Fields;
}
