/// <inheritdoc/>
public override object GetNormalisedValuePrecise(object originalValue, int identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
// Determine which textualDimensionsList to use
Dictionary <string, Dictionary <string, int> > textualDimensionsListReverse;
if (nodeDimensionData.Select(x => x.Identifier).FirstOrDefault() != null)
{
textualDimensionsListReverse = nodeTextualDimensionsListReverse;
}
else
{
textualDimensionsListReverse = edgeTextualDimensionsListReverse;
}
if (meta.Type == IATKDataType.String)
{
int stringIdx = textualDimensionsListReverse[this[identifier].Identifier][originalValue.ToString()];
return(UtilMath.NormaliseValue(stringIdx, meta.Min, meta.Max, 0f, 1f));
}
else
{
return(UtilMath.NormaliseValue((float)originalValue, meta.Min, meta.Max, 0f, 1f));
}
}
/// <summary>
/// Initialises the metadata for the data, including its string identifier and type. This function assumes that the header is the first element.
/// </summary>
/// <param name="lines">A string array representing lines from a data file.</param>
/// <returns><c>true</c> if the headers and types were successfully loaded, otherwise <c>false</c>.</returns>
private bool LoadMetaData(string[] lines, ref List <DimensionData> dimensionData)
{
if (lines.Length > 0)
{
string[] identifiers = lines[0].Split(split);
// Create metadata
DimensionData.Metadata[] metadata = new DimensionData.Metadata[identifiers.Length];
// Clean string identifiers for each dimension
for (int i = 0; i < identifiers.Length; i++)
{
string id = CleanDataString(identifiers[i]);
identifiers[i] = id;
}
// Check the types for each dimension
if (lines.Length > 1)
{
string[] typesToRead = lines[1].Split(split);
for (int i = 0; i < typesToRead.Length; i++)
{
metadata[i].Type = DataTypeExtension.inferFromString(CleanDataString(typesToRead[i]));
}
}
// Populate the dimension data list
for (int i = 0; i < identifiers.Length; ++i)
{
dimensionData.Add(new DimensionData(identifiers[i], i, metadata[i]));
}
return(true);
}
return(false);
}
/// <inheritdoc/>
public override object GetValuePrecise(float normalisedValue, string identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
// Determine which textualDimensionsList to use
Dictionary <string, Dictionary <int, string> > textualDimensionsList;
if (nodeDimensionData.FirstOrDefault(x => x.Identifier == identifier) != null)
{
textualDimensionsList = nodeTextualDimensionsList;
}
else
{
textualDimensionsList = edgeTextualDimensionsList;
}
float normValue = UtilMath.NormaliseValue(normalisedValue, 0f, 1f, meta.Min, meta.Max);
// Dimensions of type String should return a string from the textual dimensions list
if (meta.Type == IATKDataType.String)
{
return(textualDimensionsList[identifier][(int)normValue]);
}
// Otherwise re can return the de-normalised value
else
{
return(normValue);
}
}
/// <inheritdoc/>
public override object GetValueApproximate(float normalisedValue, string identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
// Determine which textualDimensionsList to use
Dictionary <string, Dictionary <int, string> > textualDimensionsList;
if (nodeDimensionData.FirstOrDefault(x => x.Identifier == identifier) != null)
{
textualDimensionsList = nodeTextualDimensionsList;
}
else
{
textualDimensionsList = edgeTextualDimensionsList;
}
// Dimensions of type String should return a string from the textual dimensions list
if (meta.Type == IATKDataType.String)
{
// Since this function allows for approximate input values, we need to find the value closest to the given one
float normValue = UtilMath.NormaliseValue(ValueClosestTo(this[identifier].Data, normalisedValue), 0f, 1f, meta.Min, meta.Max);
return(textualDimensionsList[identifier][(int)normValue]);
}
// Otherwise we can return a de-normalised value
else
{
return(UtilMath.NormaliseValue(normalisedValue, 0f, 1f, meta.Min, meta.Max));
}
}
/// <inheritdoc/>
public override object GetNormalisedValuePrecise(object originalValue, int identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
if (meta.Type == IATKDataType.String)
{
int stringIdx = textualDimensionsListReverse[this[identifier].Identifier][originalValue.ToString()];
return(UtilMath.NormaliseValue(stringIdx, meta.Min, meta.Max, 0f, 1f));
}
else
{
return(UtilMath.NormaliseValue((float)originalValue, meta.Min, meta.Max, 0f, 1f));
}
}
/// <inheritdoc/>
public override object GetValuePrecise(float normalisedValue, string identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
float normValue = UtilMath.NormaliseValue(normalisedValue, 0f, 1f, meta.Min, meta.Max);
// Dimensions of type String should return a string from the textual dimensions list
if (meta.Type == IATKDataType.String)
{
return(textualDimensionsList[this[identifier].Identifier][(int)normValue]);
}
// Otherwise re can return the de-normalised value
else
{
return(normValue);
}
}
/// <inheritdoc/>
public override object GetValueApproximate(float normalisedValue, string identifier)
{
DimensionData.Metadata meta = this[identifier].MetaData;
// Dimensions of type String should return a string from the textual dimensions list
if (meta.Type == IATKDataType.String)
{
// Since this function allows for approximate input values, we need to find the value closest to the given one
float normValue = UtilMath.NormaliseValue(ValueClosestTo(this[identifier].Data, normalisedValue), 0f, 1f, meta.Min, meta.Max);
return(textualDimensionsList[identifier][(int)normValue]);
}
// Otherwise we can return a de-normalised value
else
{
return(UtilMath.NormaliseValue(normalisedValue, 0f, 1f, meta.Min, meta.Max));
}
}
/// <summary>
/// Normalises a given column from a 2D array of float values within the range 0..1. This function also sets some metadata values.
/// </summary>
/// <param name="dataArray">A 2D float array of data.</param>
/// <param name="col">An integer index of the column to normalise.</param>
/// <returns>A normalised float array in the range 0..1.</returns>
private float[] NormaliseColumn(float[,] dataArray, int col, ref List <DimensionData> dimensionData)
{
float[] result = GetColumn(dataArray, col);
float minValue = result.Min();
float maxValue = result.Max();
if (minValue == maxValue)
{
// where there are no distinct values, need the dimension to be distinct
// otherwise lots of maths breaks with division by zero, etc.
// this is the most elegant hack I could think of, but should be fixed properly in future
minValue -= 1.0f;
maxValue += 1.0f;
}
// Populate metadata values
DimensionData.Metadata metadata = dimensionData[col].MetaData;
metadata.Min = minValue;
metadata.Max = maxValue;
metadata.Categories = result.Distinct().Select(x => UtilMath.NormaliseValue(x, minValue, maxValue, 0.0f, 1.0f)).ToArray();
metadata.CategoryCount = metadata.Categories.Count();
metadata.BinCount = (int)(maxValue - minValue + 1);
dimensionData[col].SetMetadata(metadata);
for (int j = 0; j < result.Length; j++)
{
if (minValue < maxValue)
{
result[j] = UtilMath.NormaliseValue(result[j], minValue, maxValue, 0f, 1f);
}
else
{
// Avoid NaNs or nonsensical normalization
result[j] = 0;
}
}
return(result);
}