// TODO: Merge API with an "On" parameter that merges on a column common to 2 dataframes
/// <summary>
/// Merge DataFrames with a database style join
/// </summary>
/// <param name="other"></param>
/// <param name="leftJoinColumn"></param>
/// <param name="rightJoinColumn"></param>
/// <param name="leftSuffix"></param>
/// <param name="rightSuffix"></param>
/// <param name="joinAlgorithm"></param>
/// <returns></returns>
public DataFrame Merge <TKey>(DataFrame other, string leftJoinColumn, string rightJoinColumn, string leftSuffix = "_left", string rightSuffix = "_right", JoinAlgorithm joinAlgorithm = JoinAlgorithm.Left)
{
// A simple hash join
DataFrame ret = new DataFrame();
DataFrame leftDataFrame = this;
DataFrame rightDataFrame = other;
// The final table size is not known until runtime
long rowNumber = 0;
PrimitiveDataFrameColumn <long> leftRowIndices = new PrimitiveDataFrameColumn <long>("LeftIndices");
PrimitiveDataFrameColumn <long> rightRowIndices = new PrimitiveDataFrameColumn <long>("RightIndices");
if (joinAlgorithm == JoinAlgorithm.Left)
{
// First hash other dataframe on the rightJoinColumn
DataFrameColumn otherColumn = other[rightJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = otherColumn.GroupColumnValues <TKey>();
// Go over the records in this dataframe and match with the dictionary
DataFrameColumn thisColumn = this[leftJoinColumn];
for (long i = 0; i < thisColumn.Length; i++)
{
var thisColumnValue = thisColumn[i];
TKey thisColumnValueOrDefault = (TKey)(thisColumnValue == null ? default(TKey) : thisColumnValue);
if (multimap.TryGetValue(thisColumnValueOrDefault, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
if (thisColumnValue == null)
{
// Match only with nulls in otherColumn
if (otherColumn[row] == null)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
}
}
else
{
// Cannot match nulls in otherColumn
if (otherColumn[row] != null)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
}
}
}
}
else
{
leftRowIndices.Append(i);
rightRowIndices.Append(null);
}
}
}
else if (joinAlgorithm == JoinAlgorithm.Right)
{
DataFrameColumn thisColumn = this[leftJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = thisColumn.GroupColumnValues <TKey>();
DataFrameColumn otherColumn = other[rightJoinColumn];
for (long i = 0; i < otherColumn.Length; i++)
{
var otherColumnValue = otherColumn[i];
TKey otherColumnValueOrDefault = (TKey)(otherColumnValue == null ? default(TKey) : otherColumnValue);
if (multimap.TryGetValue(otherColumnValueOrDefault, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
if (otherColumnValue == null)
{
if (thisColumn[row] == null)
{
leftRowIndices.Append(row);
rightRowIndices.Append(i);
}
}
else
{
if (thisColumn[row] != null)
{
leftRowIndices.Append(row);
rightRowIndices.Append(i);
}
}
}
}
else
{
leftRowIndices.Append(null);
rightRowIndices.Append(i);
}
}
}
else if (joinAlgorithm == JoinAlgorithm.Inner)
{
// Hash the column with the smaller RowCount
long leftRowCount = Rows.Count;
long rightRowCount = other.Rows.Count;
DataFrame longerDataFrame = leftRowCount <= rightRowCount ? other : this;
DataFrame shorterDataFrame = ReferenceEquals(longerDataFrame, this) ? other : this;
DataFrameColumn hashColumn = (leftRowCount <= rightRowCount) ? this[leftJoinColumn] : other[rightJoinColumn];
DataFrameColumn otherColumn = ReferenceEquals(hashColumn, this[leftJoinColumn]) ? other[rightJoinColumn] : this[leftJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = hashColumn.GroupColumnValues <TKey>();
for (long i = 0; i < otherColumn.Length; i++)
{
var otherColumnValue = otherColumn[i];
TKey otherColumnValueOrDefault = (TKey)(otherColumnValue == null ? default(TKey) : otherColumnValue);
if (multimap.TryGetValue(otherColumnValueOrDefault, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
if (otherColumnValue == null)
{
if (hashColumn[row] == null)
{
leftRowIndices.Append(row);
rightRowIndices.Append(i);
}
}
else
{
if (hashColumn[row] != null)
{
leftRowIndices.Append(row);
rightRowIndices.Append(i);
}
}
}
}
}
leftDataFrame = shorterDataFrame;
rightDataFrame = longerDataFrame;
}
else if (joinAlgorithm == JoinAlgorithm.FullOuter)
{
DataFrameColumn otherColumn = other[rightJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = otherColumn.GroupColumnValues <TKey>();
Dictionary <TKey, long> intersection = new Dictionary <TKey, long>(EqualityComparer <TKey> .Default);
// Go over the records in this dataframe and match with the dictionary
DataFrameColumn thisColumn = this[leftJoinColumn];
for (long i = 0; i < thisColumn.Length; i++)
{
var thisColumnValue = thisColumn[i];
TKey thisColumnValueOrDefault = (TKey)(thisColumnValue == null ? default(TKey) : thisColumnValue);
if (multimap.TryGetValue(thisColumnValueOrDefault, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
if (thisColumnValue == null)
{
// Has to match only with nulls in otherColumn
if (otherColumn[row] == null)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
if (!intersection.ContainsKey(thisColumnValueOrDefault))
{
intersection.Add(thisColumnValueOrDefault, rowNumber);
}
}
}
else
{
// Cannot match to nulls in otherColumn
if (otherColumn[row] != null)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
if (!intersection.ContainsKey(thisColumnValueOrDefault))
{
intersection.Add(thisColumnValueOrDefault, rowNumber);
}
}
}
}
}
else
{
leftRowIndices.Append(i);
rightRowIndices.Append(null);
}
}
for (long i = 0; i < otherColumn.Length; i++)
{
TKey value = (TKey)(otherColumn[i] ?? default(TKey));
if (!intersection.ContainsKey(value))
{
leftRowIndices.Append(null);
rightRowIndices.Append(i);
}
}
}
else
{
throw new NotImplementedException(nameof(joinAlgorithm));
}
for (int i = 0; i < leftDataFrame.Columns.Count; i++)
{
ret.Columns.Insert(i, leftDataFrame.Columns[i].Clone(leftRowIndices));
}
for (int i = 0; i < rightDataFrame.Columns.Count; i++)
{
DataFrameColumn column = rightDataFrame.Columns[i].Clone(rightRowIndices);
SetSuffixForDuplicatedColumnNames(ret, column, leftSuffix, rightSuffix);
ret.Columns.Insert(ret.Columns.Count, column);
}
return(ret);
}
// TODO: Merge API with an "On" parameter that merges on a column common to 2 dataframes
/// <summary>
/// Merge DataFrames with a database style join
/// </summary>
/// <param name="other"></param>
/// <param name="leftJoinColumn"></param>
/// <param name="rightJoinColumn"></param>
/// <param name="leftSuffix"></param>
/// <param name="rightSuffix"></param>
/// <param name="joinAlgorithm"></param>
/// <returns></returns>
public DataFrame Merge <TKey>(DataFrame other, string leftJoinColumn, string rightJoinColumn, string leftSuffix = "_left", string rightSuffix = "_right", JoinAlgorithm joinAlgorithm = JoinAlgorithm.Left)
{
// A simple hash join
DataFrame ret = new DataFrame();
DataFrame leftDataFrame = this;
DataFrame rightDataFrame = other;
// The final table size is not known until runtime
long rowNumber = 0;
PrimitiveDataFrameColumn <long> leftRowIndices = new PrimitiveDataFrameColumn <long>("LeftIndices");
PrimitiveDataFrameColumn <long> rightRowIndices = new PrimitiveDataFrameColumn <long>("RightIndices");
if (joinAlgorithm == JoinAlgorithm.Left)
{
// First hash other dataframe on the rightJoinColumn
DataFrameColumn otherColumn = other.Columns[rightJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = otherColumn.GroupColumnValues <TKey>(out HashSet <long> otherColumnNullIndices);
// Go over the records in this dataframe and match with the dictionary
DataFrameColumn thisColumn = Columns[leftJoinColumn];
for (long i = 0; i < thisColumn.Length; i++)
{
var thisColumnValue = thisColumn[i];
if (thisColumnValue != null)
{
if (multimap.TryGetValue((TKey)thisColumnValue, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
}
}
else
{
leftRowIndices.Append(i);
rightRowIndices.Append(null);
}
}
else
{
foreach (long row in otherColumnNullIndices)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
}
}
}
}
else if (joinAlgorithm == JoinAlgorithm.Right)
{
DataFrameColumn thisColumn = Columns[leftJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = thisColumn.GroupColumnValues <TKey>(out HashSet <long> thisColumnNullIndices);
DataFrameColumn otherColumn = other.Columns[rightJoinColumn];
for (long i = 0; i < otherColumn.Length; i++)
{
var otherColumnValue = otherColumn[i];
if (otherColumnValue != null)
{
if (multimap.TryGetValue((TKey)otherColumnValue, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
leftRowIndices.Append(row);
rightRowIndices.Append(i);
}
}
else
{
leftRowIndices.Append(null);
rightRowIndices.Append(i);
}
}
else
{
foreach (long thisColumnNullIndex in thisColumnNullIndices)
{
leftRowIndices.Append(thisColumnNullIndex);
rightRowIndices.Append(i);
}
}
}
}
else if (joinAlgorithm == JoinAlgorithm.Inner)
{
// Hash the column with the smaller RowCount
long leftRowCount = Rows.Count;
long rightRowCount = other.Rows.Count;
bool leftColumnIsSmaller = leftRowCount <= rightRowCount;
DataFrameColumn hashColumn = leftColumnIsSmaller ? Columns[leftJoinColumn] : other.Columns[rightJoinColumn];
DataFrameColumn otherColumn = ReferenceEquals(hashColumn, Columns[leftJoinColumn]) ? other.Columns[rightJoinColumn] : Columns[leftJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = hashColumn.GroupColumnValues <TKey>(out HashSet <long> smallerDataFrameColumnNullIndices);
for (long i = 0; i < otherColumn.Length; i++)
{
var otherColumnValue = otherColumn[i];
if (otherColumnValue != null)
{
if (multimap.TryGetValue((TKey)otherColumnValue, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
leftRowIndices.Append(leftColumnIsSmaller ? row : i);
rightRowIndices.Append(leftColumnIsSmaller ? i : row);
}
}
}
else
{
foreach (long nullIndex in smallerDataFrameColumnNullIndices)
{
leftRowIndices.Append(leftColumnIsSmaller ? nullIndex : i);
rightRowIndices.Append(leftColumnIsSmaller ? i : nullIndex);
}
}
}
}
else if (joinAlgorithm == JoinAlgorithm.FullOuter)
{
DataFrameColumn otherColumn = other.Columns[rightJoinColumn];
Dictionary <TKey, ICollection <long> > multimap = otherColumn.GroupColumnValues <TKey>(out HashSet <long> otherColumnNullIndices);
Dictionary <TKey, long> intersection = new Dictionary <TKey, long>(EqualityComparer <TKey> .Default);
// Go over the records in this dataframe and match with the dictionary
DataFrameColumn thisColumn = Columns[leftJoinColumn];
Int64DataFrameColumn thisColumnNullIndices = new Int64DataFrameColumn("ThisColumnNullIndices");
for (long i = 0; i < thisColumn.Length; i++)
{
var thisColumnValue = thisColumn[i];
if (thisColumnValue != null)
{
if (multimap.TryGetValue((TKey)thisColumnValue, out ICollection <long> rowNumbers))
{
foreach (long row in rowNumbers)
{
leftRowIndices.Append(i);
rightRowIndices.Append(row);
if (!intersection.ContainsKey((TKey)thisColumnValue))
{
intersection.Add((TKey)thisColumnValue, rowNumber);
}
}
}
else
{
leftRowIndices.Append(i);
rightRowIndices.Append(null);
}
}
else
{
thisColumnNullIndices.Append(i);
}
}
for (long i = 0; i < otherColumn.Length; i++)
{
var value = otherColumn[i];
if (value != null)
{
if (!intersection.ContainsKey((TKey)value))
{
leftRowIndices.Append(null);
rightRowIndices.Append(i);
}
}
}
// Now handle the null rows
foreach (long?thisColumnNullIndex in thisColumnNullIndices)
{
foreach (long otherColumnNullIndex in otherColumnNullIndices)
{
leftRowIndices.Append(thisColumnNullIndex.Value);
rightRowIndices.Append(otherColumnNullIndex);
}
if (otherColumnNullIndices.Count == 0)
{
leftRowIndices.Append(thisColumnNullIndex.Value);
rightRowIndices.Append(null);
}
}
if (thisColumnNullIndices.Length == 0)
{
foreach (long otherColumnNullIndex in otherColumnNullIndices)
{
leftRowIndices.Append(null);
rightRowIndices.Append(otherColumnNullIndex);
}
}
}
else
{
throw new NotImplementedException(nameof(joinAlgorithm));
}
for (int i = 0; i < leftDataFrame.Columns.Count; i++)
{
ret.Columns.Insert(i, leftDataFrame.Columns[i].Clone(leftRowIndices));
}
for (int i = 0; i < rightDataFrame.Columns.Count; i++)
{
DataFrameColumn column = rightDataFrame.Columns[i].Clone(rightRowIndices);
SetSuffixForDuplicatedColumnNames(ret, column, leftSuffix, rightSuffix);
ret.Columns.Insert(ret.Columns.Count, column);
}
return(ret);
}
