internal override IEnumerable <Row> GetEntities(DataExtractorHelper schema)
{
foreach (var record in page.Records)
{
var dataRow = schema.NewRow();
var readState = new RecordReadState(schema.Columns.Count(x => x.UnderlyingType == ColumnType.Bit));
int columnIndex = 0;
foreach (DataColumn col in dataRow.Columns)
{
var sqlType = SqlTypeFactory.Create(col, readState, new CompressionContext(CompressionLevel.Row, true));
IVariableLengthDataProxy dataProxy = record.GetPhysicalColumnBytes(columnIndex);
if (dataProxy == null)
{
dataRow[col] = null;
}
else
{
dataRow[col] = sqlType.GetValue(dataProxy.GetBytes().ToArray());
}
columnIndex++;
}
yield return(dataRow);
}
}
/// <summary>
/// Starts at the data page (loc) and follows the NextPage pointer chain till the end.
/// </summary>
internal IEnumerable <Row> ScanLinkedDataPages(PagePointer loc, DataExtractorHelper schema,
CompressionContext compression, bool isSysTable)
{
while (PagePointer.Zero != loc && loc != null && loc.PageID > 0)
{
var recordParser = RecordEntityParser.CreateEntityParserForPage(loc, compression, Database, isSysTable);
foreach (var dr in recordParser.GetEntities(schema))
{
yield return(dr);
}
loc = recordParser.NextPage;
}
}
internal override IEnumerable <Row> GetEntities(DataExtractorHelper schema)
{
foreach (var record in page.Records)
{
// Don't process forwarded blob fragments as they should only be processed from the referenced record
if (_recordsToSkip.Contains(record.Type) || record.IsGhostForwardedRecord)
{
continue;
}
short fixedOffset = 0;
short variableColumnIndex = 0;
var dataRow = schema.NewRow();
var readState = new RecordReadState(schema.BitColumnsCount);
var bitColumnBytes = new byte[0];
foreach (var col in schema.Columns)
{
var sqlType = SqlTypeFactory.Create(col, readState, compression);
object columnValue = null;
// Sparse columns needs to retrieve their values from the sparse vector, contained in the very last
// variable length column in the record.
if (col.IsSparse)
{
// We may encounter records that don't have any sparse vectors, for instance if no sparse columns have values
if (record.SparseVector != null)
{
// Column ID's are stored as ints in general. In the sparse vector though, they're stored as shorts.
if (record.SparseVector.ColumnValues.ContainsKey((short)col.ColumnID))
{
columnValue = sqlType.GetValue(record.SparseVector.ColumnValues[(short)col.ColumnID]);
}
}
}
else
{
var nonSparseIndex = schema.NonSparseIndexes[col.Name];
// Before we even try to parse the column & make a null bitmap lookup, ensure that it's present in the record.
// There may be columns > record.NumberOfColumns caused by nullable columns added to the schema after the record was written.
if (nonSparseIndex < record.NumberOfColumns && col.UnderlyingType != ColumnType.Computed)
{
if (sqlType.IsVariableLength)
{
// If there's either no null bitmap, or the null bitmap defines the column as non-null.
if (!record.HasNullBitmap || !record.NullBitmap[nonSparseIndex])
{
// If the current variable length column index exceeds the number of stored
// variable length columns, the value is empty by definition (that is, 0 bytes, but not null).
if (variableColumnIndex < record.NumberOfVariableLengthColumns)
{
var data = record.VariableLengthColumnData[variableColumnIndex].GetBytes()?.ToArray();
columnValue = sqlType.GetValue(data ?? new byte[0]);
}
else
{
columnValue = sqlType.GetValue(new byte[0]);
}
}
variableColumnIndex++;
}
else
{
// Must cache type FixedLength as it may change after getting a value (e.g. SqlBit)
var fixedLength = sqlType.FixedLength.Value;
if ((!record.HasNullBitmap || !record.NullBitmap[nonSparseIndex]) && col.UnderlyingType != ColumnType.Bit)
{
var valueBytes = record.FixedLengthData.Skip(fixedOffset).Take(fixedLength).ToArray();
// We may run out of fixed length bytes. In certain conditions a null integer may have been added without
// there being a null bitmap. In such a case, we detect the null condition by there not being enough fixed
// length bytes to process.
if (valueBytes.Length > 0)
{
columnValue = sqlType.GetValue(valueBytes);
}
}
else if (col.UnderlyingType == ColumnType.Bit && !schema.IsDroppedColumn(col))
{
if (readState.IsFirstBit)
{
bitColumnBytes = record.FixedLengthData.Skip(fixedOffset).Take(fixedLength).ToArray();
}
var value = sqlType.GetValue(bitColumnBytes);
columnValue = !record.HasNullBitmap || !record.NullBitmap[nonSparseIndex] ? value : null;
}
fixedOffset += fixedLength;
}
}
else if (col.UnderlyingType == ColumnType.Computed)
{
columnValue = sqlType.GetValue(null);
}
else if (!col.IsNullable)
{
columnValue = schema.GetDefaultValue(col, sqlType);
}
}
if (!schema.IsDroppedColumn(col))
{
dataRow[col] = columnValue;
}
}
yield return(dataRow);
}
}
private IEnumerable <Row> ScanPartition(long partitionID, int partitionNumber, Row schema, bool isSysTable = true)
{
// Lookup partition
var partition = Database.Dmvs.Partitions
.SingleOrDefault(p => p.PartitionID == partitionID && p.PartitionNumber == partitionNumber);
if (partition == null)
{
throw new ArgumentException("Partition (" + partitionID + "." + partitionNumber + " does not exist.");
}
// Get allocation unit for in-row data
var au = Database.Dmvs.SystemInternalsAllocationUnits
.SingleOrDefault(x => x.ContainerID == partition.PartitionID && x.Type == 1);
if (au == null)
{
throw new ArgumentException("Partition (" + partition.PartitionID + "." + partition.PartitionNumber + " has no HOBT allocation unit.");
}
// Before we can scan either heaps or indices, we need to know the compression level as that's set at the partition level, and not at the record/page level.
// We also need to know whether the partition is using vardecimals.
var compression = new CompressionContext((CompressionLevel)partition.DataCompression, MetaData.PartitionHasVardecimalColumns(partition.PartitionID));
var clusteredIndex = isSysTable ? null : Database.Dmvs.Indexes.SingleOrDefault(x => x.ObjectID == partition.ObjectID && x.Type == 1);
var useClusteredIndex = isSysTable || clusteredIndex != null;
var partitionColumns = isSysTable ? null : Database.Dmvs.SystemInternalsPartitionColumns.Where(x => x.PartitionID == partition.PartitionID).ToArray();
var defaultConstraints = isSysTable ? null : Database.Dmvs.SysDefaultConstraints.Where(x => x.ParentObjectId == partition.ObjectID).ToArray();
var schemaWrapper = new DataExtractorHelper(schema, Database.Dmvs, null, partitionColumns, defaultConstraints);
// Heap tables won't have root pages, thus we can check whether a root page is defined for the HOBT allocation unit
if (au.RootPagePointer != PagePointer.Zero && useClusteredIndex)
{
var currentPage = isSysTable ? au.FirstPagePointer : au.RootPagePointer;
if (currentPage != au.FirstPagePointer)
{
while (true)
{
var ciPage = Database.GetClusteredIndexPage(currentPage, isSysTable);
currentPage = ciPage.Records.Select(x => x.PageId).FirstOrDefault();
if (ciPage.Header.Level <= 1)
{
break;
}
}
}
// Index
foreach (var row in ScanLinkedDataPages(currentPage, schemaWrapper, compression, isSysTable))
{
yield return(row);
}
}
else
{
// Heap
foreach (var row in ScanHeap(au.FirstIamPagePointer, schemaWrapper, compression, isSysTable))
{
yield return(row);
}
}
}
/// <summary>
/// Scans a heap beginning from the provided IAM page and onwards.
/// </summary>
private IEnumerable <Row> ScanHeap(PagePointer loc, DataExtractorHelper schema, CompressionContext compression,
bool isSysTable)
{
// Traverse the linked list of IAM pages until the tail pointer is zero
while (loc != PagePointer.Zero)
{
// Before scanning, check that the IAM page itself is allocated
var pfsPage = Database.GetPfsPage(PfsPage.GetPfsPointerForPage(loc));
// If IAM page isn't allocated, there's nothing to return
if (!pfsPage.GetPageDescription(loc.PageID).IsAllocated)
{
yield break;
}
var iamPage = Database.GetIamPage(loc, isSysTable);
// Create an array with all of the header slot pointers
var iamPageSlots = new []
{
iamPage.Slot0,
iamPage.Slot1,
iamPage.Slot2,
iamPage.Slot3,
iamPage.Slot4,
iamPage.Slot5,
iamPage.Slot6,
iamPage.Slot7
};
// Loop each header slot and yield the results, provided the header slot is allocated
foreach (var slot in iamPageSlots.Where(x => x != PagePointer.Zero))
{
var recordParser = RecordEntityParser.CreateEntityParserForPage(slot, compression, Database, isSysTable);
foreach (var dr in recordParser.GetEntities(schema))
{
yield return(dr);
}
}
// Then loop through allocated extents and yield results
foreach (var extent in iamPage.GetAllocatedExtents())
{
// Get PFS page that tracks this extent
var pfs = Database.GetPfsPage(PfsPage.GetPfsPointerForPage(extent.StartPage));
foreach (var pageLoc in extent.GetPagePointers())
{
// Check if page is allocated according to PFS page
var pfsDescription = pfs.GetPageDescription(pageLoc.PageID);
if (!pfsDescription.IsAllocated)
{
continue;
}
var recordParser = RecordEntityParser.CreateEntityParserForPage(pageLoc, compression, Database, isSysTable);
foreach (var dr in recordParser.GetEntities(schema))
{
yield return(dr);
}
}
}
// Update current IAM chain location to the tail pointer
loc = iamPage.Header.NextPage;
}
}
internal abstract IEnumerable <Row> GetEntities(DataExtractorHelper schema);