/// <summary>
/// Describes how the transformer handles one column pair.
/// </summary>
/// <param name="name">Name of the column resulting from the transformation of <paramref name="inputColumnName"/>.</param>
/// <param name="inputColumnName">Name of column to transform. If set to <see langword="null"/>, the value of the <paramref name="name"/> will be used as source.</param>
/// <param name="outputKind">Kind of output: bag, indicator vector etc.</param>
/// <param name="hashBits">Number of bits to hash into. Must be between 1 and 31, inclusive.</param>
/// <param name="seed">Hashing seed.</param>
/// <param name="ordered">Whether the position of each term should be included in the hash.</param>
/// <param name="invertHash">During hashing we constuct mappings between original values and the produced hash values.
/// Text representation of original values are stored in the slot names of the metadata for the new column.Hashing, as such, can map many initial values to one.
/// <paramref name="invertHash"/> specifies the upper bound of the number of distinct input values mapping to a hash that should be retained.
/// <value>0</value> does not retain any input values. <value>-1</value> retains all input values mapping to each hash.</param>
public ColumnInfo(string name, string inputColumnName = null,
OneHotEncodingTransformer.OutputKind outputKind = Defaults.OutputKind,
int hashBits = Defaults.HashBits,
uint seed = Defaults.Seed,
bool ordered = Defaults.Ordered,
int invertHash = Defaults.InvertHash)
{
HashInfo = new HashingEstimator.ColumnInfo(name, inputColumnName ?? name, hashBits, seed, ordered, invertHash);
OutputKind = outputKind;
}
/// <summary>
/// Ensures the provided <paramref name="samplingKeyColumn"/> is valid for <see cref="RangeFilter"/>, hashing it if necessary, or creates a new column <paramref name="samplingKeyColumn"/> is null.
/// </summary>
private void EnsureGroupPreservationColumn(ref IDataView data, ref string samplingKeyColumn, uint?seed = null)
{
// We need to handle two cases: if samplingKeyColumn is provided, we use hashJoin to
// build a single hash of it. If it is not, we generate a random number.
if (samplingKeyColumn == null)
{
samplingKeyColumn = data.Schema.GetTempColumnName("IdPreservationColumn");
data = new GenerateNumberTransform(Environment, data, samplingKeyColumn, seed);
}
else
{
if (!data.Schema.TryGetColumnIndex(samplingKeyColumn, out int stratCol))
{
throw Environment.ExceptSchemaMismatch(nameof(samplingKeyColumn), "GroupPreservationColumn", samplingKeyColumn);
}
var type = data.Schema[stratCol].Type;
if (!RangeFilter.IsValidRangeFilterColumnType(Environment, type))
{
// Hash the samplingKeyColumn.
// REVIEW: this could currently crash, since Hash only accepts a limited set
// of column types. It used to be HashJoin, but we should probably extend Hash
// instead of having two hash transformations.
var origStratCol = samplingKeyColumn;
int tmp;
int inc = 0;
// Generate a new column with the hashed samplingKeyColumn.
while (data.Schema.TryGetColumnIndex(samplingKeyColumn, out tmp))
{
samplingKeyColumn = string.Format("{0}_{1:000}", origStratCol, ++inc);
}
HashingEstimator.ColumnInfo columnInfo;
if (seed.HasValue)
{
columnInfo = new HashingEstimator.ColumnInfo(samplingKeyColumn, origStratCol, 30, seed.Value);
}
else
{
columnInfo = new HashingEstimator.ColumnInfo(samplingKeyColumn, origStratCol, 30);
}
data = new HashingEstimator(Environment, columnInfo).Fit(data).Transform(data);
}
}
}
private void InitMap <T>(T val, ColumnType type, int hashBits = 20, ValueGetter <T> getter = null)
{
if (getter == null)
{
getter = (ref T dst) => dst = val;
}
_inRow = RowImpl.Create(type, getter);
// One million features is a nice, typical number.
var info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: hashBits);
var xf = new HashingTransformer(_env, new[] { info });
var mapper = xf.GetRowToRowMapper(_inRow.Schema);
var column = mapper.OutputSchema["Bar"];
var outRow = mapper.GetRow(_inRow, c => c == column.Index);
if (type is VectorType)
{
_vecGetter = outRow.GetGetter <VBuffer <uint> >(column.Index);
}
else
{
_getter = outRow.GetGetter <uint>(column.Index);
}
}
private void HashTestCore <T>(T val, PrimitiveType type, uint expected, uint expectedOrdered, uint expectedOrdered3)
{
const int bits = 10;
var builder = new MetadataBuilder();
builder.AddPrimitiveValue("Foo", type, val);
var inRow = MetadataUtils.MetadataAsRow(builder.GetMetadata());
// First do an unordered hash.
var info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits);
var xf = new HashingTransformer(Env, new[] { info });
var mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out int outCol);
var outRow = mapper.GetRow(inRow, c => c == outCol);
var getter = outRow.GetGetter <uint>(outCol);
uint result = 0;
getter(ref result);
Assert.Equal(expected, result);
// Next do an ordered hash.
info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits, ordered: true);
xf = new HashingTransformer(Env, new[] { info });
mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out outCol);
outRow = mapper.GetRow(inRow, c => c == outCol);
getter = outRow.GetGetter <uint>(outCol);
getter(ref result);
Assert.Equal(expectedOrdered, result);
// Next build up a vector to make sure that hashing is consistent between scalar values
// at least in the first position, and in the unordered case, the last position.
const int vecLen = 5;
var denseVec = new VBuffer <T>(vecLen, Utils.CreateArray(vecLen, val));
builder = new MetadataBuilder();
builder.Add("Foo", new VectorType(type, vecLen), (ref VBuffer <T> dst) => denseVec.CopyTo(ref dst));
inRow = MetadataUtils.MetadataAsRow(builder.GetMetadata());
info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits, ordered: false);
xf = new HashingTransformer(Env, new[] { info });
mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out outCol);
outRow = mapper.GetRow(inRow, c => c == outCol);
var vecGetter = outRow.GetGetter <VBuffer <uint> >(outCol);
VBuffer <uint> vecResult = default;
vecGetter(ref vecResult);
Assert.Equal(vecLen, vecResult.Length);
// They all should equal this in this case.
Assert.All(vecResult.DenseValues(), v => Assert.Equal(expected, v));
// Now do ordered with the dense vector.
info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits, ordered: true);
xf = new HashingTransformer(Env, new[] { info });
mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out outCol);
outRow = mapper.GetRow(inRow, c => c == outCol);
vecGetter = outRow.GetGetter <VBuffer <uint> >(outCol);
vecGetter(ref vecResult);
Assert.Equal(vecLen, vecResult.Length);
Assert.Equal(expectedOrdered, vecResult.GetItemOrDefault(0));
Assert.Equal(expectedOrdered3, vecResult.GetItemOrDefault(3));
Assert.All(vecResult.DenseValues(), v => Assert.True((v == 0) == (expectedOrdered == 0)));
// Let's now do a sparse vector.
var sparseVec = new VBuffer <T>(10, 3, Utils.CreateArray(3, val), new[] { 0, 3, 7 });
builder = new MetadataBuilder();
builder.Add("Foo", new VectorType(type, vecLen), (ref VBuffer <T> dst) => sparseVec.CopyTo(ref dst));
inRow = MetadataUtils.MetadataAsRow(builder.GetMetadata());
info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits, ordered: false);
xf = new HashingTransformer(Env, new[] { info });
mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out outCol);
outRow = mapper.GetRow(inRow, c => c == outCol);
vecGetter = outRow.GetGetter <VBuffer <uint> >(outCol);
vecGetter(ref vecResult);
Assert.Equal(10, vecResult.Length);
Assert.Equal(expected, vecResult.GetItemOrDefault(0));
Assert.Equal(expected, vecResult.GetItemOrDefault(3));
Assert.Equal(expected, vecResult.GetItemOrDefault(7));
info = new HashingEstimator.ColumnInfo("Bar", "Foo", hashBits: bits, ordered: true);
xf = new HashingTransformer(Env, new[] { info });
mapper = xf.GetRowToRowMapper(inRow.Schema);
mapper.OutputSchema.TryGetColumnIndex("Bar", out outCol);
outRow = mapper.GetRow(inRow, c => c == outCol);
vecGetter = outRow.GetGetter <VBuffer <uint> >(outCol);
vecGetter(ref vecResult);
Assert.Equal(10, vecResult.Length);
Assert.Equal(expectedOrdered, vecResult.GetItemOrDefault(0));
Assert.Equal(expectedOrdered3, vecResult.GetItemOrDefault(3));
}
