public DataViewRowCursor GetRowCursor(IEnumerable <DataViewSchema.Column> columnsNeeded, Random rand = null)
{
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, Schema);
_host.CheckValueOrNull(rand);
var srcPredicates = _zipBinding.GetInputPredicates(predicate);
// REVIEW: if we know the row counts, we could only open cursor if it has needed columns, and have the
// outer cursor handle the early stopping. If we don't know row counts, we need to open all the cursors because
// we don't know which one will be the shortest.
// One reason this is not done currently is because the API has 'somewhat mutable' data views, so potentially this
// optimization might backfire.
var srcCursors = _sources
.Select((dv, i) => srcPredicates[i] == null ? GetMinimumCursor(dv) : dv.GetRowCursor(dv.Schema.Where(x => srcPredicates[i](x.Index)), null)).ToArray();
return(new Cursor(this, srcCursors, predicate));
}
public DataViewRowCursor GetRowCursor(IEnumerable <DataViewSchema.Column> columnsNeeded, Random rand = null)
{
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, OutputSchema);
Host.CheckValueOrNull(rand);
// If we aren't selecting any of the output columns, don't construct our cursor.
// Note that because we cannot support random due to the inherently
// stratified nature, neither can we allow the base data to be shuffled,
// even if it supports shuffling.
var bindings = GetBindings();
if (!bindings.AnyNewColumnsActive(predicate))
{
var activeInput = bindings.GetActiveInput(predicate);
var activeCols = Source.Schema.Where(x => activeInput.Length > x.Index && activeInput[x.Index]);
var inputCursor = Source.GetRowCursor(activeCols, null);
return(new BindingsWrappedRowCursor(Host, inputCursor, bindings));
}
return(GetRowCursorCore(predicate));
}
public override DataViewRowCursor[] GetRowCursorSet(IEnumerable <DataViewSchema.Column> columnsNeeded, int n, Random rand = null)
{
Host.CheckValueOrNull(rand);
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, OutputSchema);
bool[] active;
Func <int, bool> inputPred = GetActive(predicate, out active);
var inputCols = Source.Schema.Where(x => inputPred(x.Index));
var inputs = Source.GetRowCursorSet(inputCols, n, rand);
Host.AssertNonEmpty(inputs);
// No need to split if this is given 1 input cursor.
var cursors = new DataViewRowCursor[inputs.Length];
for (int i = 0; i < inputs.Length; i++)
{
cursors[i] = new Cursor(this, inputs[i], active);
}
return(cursors);
}
public sealed override RowCursor[] GetRowCursorSet(IEnumerable<Schema.Column> columnsNeeded, int n, Random rand = null)
{
Host.CheckValueOrNull(rand);
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, OutputSchema);
var inputPred = _bindings.GetDependencies(predicate);
var active = _bindings.GetActive(predicate);
var inputCols = Source.Schema.Where(x => inputPred(x.Index));
var inputs = Source.GetRowCursorSet(inputCols, n, rand);
Host.AssertNonEmpty(inputs);
if (inputs.Length == 1 && n > 1 && WantParallelCursors(predicate))
inputs = DataViewUtils.CreateSplitCursors(Host, inputs[0], n);
Host.AssertNonEmpty(inputs);
var cursors = new RowCursor[inputs.Length];
for (int i = 0; i < inputs.Length; i++)
cursors[i] = new Cursor(Host, this, inputs[i], active);
return cursors;
}
public RowCursor GetRowCursor(IEnumerable<Schema.Column> columnsNeeded, Random rand = null)
{
Host.CheckValueOrNull(rand);
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, OutputSchema);
var rng = CanShuffle ? rand : null;
bool? useParallel = ShouldUseParallelCursors(predicate);
// When useParallel is null, let the input decide, so go ahead and ask for parallel.
// When the input wants to be split, this puts the consolidation after this transform
// instead of before. This is likely to produce better performance, for example, when
// this is RangeFilter.
RowCursor curs;
if (useParallel != false &&
DataViewUtils.TryCreateConsolidatingCursor(out curs, this, columnsNeeded, Host, rng))
{
return curs;
}
return GetRowCursorCore(columnsNeeded, rng);
}
public sealed override RowCursor[] GetRowCursorSet(IEnumerable <Schema.Column> columnsNeeded, int n, Random rand = null)
{
Host.CheckValueOrNull(rand);
var predicate = RowCursorUtils.FromColumnsToPredicate(columnsNeeded, OutputSchema);
var inputPred = _bindings.GetDependencies(predicate);
var active = _bindings.GetActive(predicate);
var inputCols = Source.Schema.Where(x => inputPred(x.Index));
var inputs = Source.GetRowCursorSet(inputCols, n, rand);
Host.AssertNonEmpty(inputs);
// No need to split if this is given 1 input cursor.
var cursors = new RowCursor[inputs.Length];
for (int i = 0; i < inputs.Length; i++)
{
cursors[i] = new Cursor(Host, _bindings, inputs[i], active);
}
return(cursors);
}