/// <summary>
/// Read an individual line from a serialized probability hash
/// into a single Markov node.
/// </summary>
/// <param name="data">Parsing state data</param>
private static void ParseLine(CorpusParseData data)
{
// Line is split into two parts (key and value)
var components = data.Line.Split(CorpusWriter.ListDelimiter);
if (components.Length != 2)
throw new InvalidDataException("Two list delimiters on a line");
// The key is split in two parts (list of key words and a sub-topic)
var keyComponents = components[0].Split(CorpusWriter.TopicDelimiter);
if (keyComponents.Length != 2)
throw new InvalidDataException("Two topic delimiters in a key");
// There are n keywords for an n-order Markov chain
var keyWords = keyComponents[0].Split(CorpusWriter.ItemDelimiter);
if (keyWords.Length != data.ProbabilityHash.Order)
throw new InvalidDataException("Invalid number of keywords");
var topic = keyComponents[1];
// Get the list of words which follow the key in the corpus, including
// their probabilities.
var valueItems = components[1].Split(CorpusWriter.ItemDelimiter);
var key = new ProbabilityHashKey(keyWords.ToList(), topic);
// Finalize the value because the probabilities are already
// present in the value and don't need to be calculated later
var value = new ProbabilityHashValue();
value.Finalize();
foreach (var valueItem in valueItems)
{
// Separate each word from its probability
var valueComponents = valueItem.Split(CorpusWriter.ProbabilityDelimiter);
if (valueComponents.Length != 2)
throw new InvalidDataException("Multiple probabilities for a word");
// Ensure the probability is formatted
double probability;
if (!Double.TryParse(valueComponents[1], out probability))
throw new InvalidDataException("Probability is not a real number");
value.Add(valueComponents[0], probability);
}
data.ProbabilityHash.Add(key, value);
}
/// <summary>
/// Parses a serialized probability hash from disk.
///
/// Throws a FileNotFound exception if the specified data doesn't yet exist.
/// Throws an InvalidDataException if the data format is invalid.
/// </summary>
/// <param name="name">Major topic name</param>
/// <param name="order">Markov order</param>
/// <returns></returns>
public static ProbabilityHash Parse(string name, int order)
{
// Mangle the filenames to not contain spaces
name = name.Replace(" ", "-");
// Probability hash is finalized because word count data is already
// present in the file, and doesn't need to be recalculated.
var probabilityHash = new ProbabilityHash(order);
probabilityHash.Finalize();
var data = new CorpusParseData()
{
ProbabilityHash = probabilityHash,
};
// This was a rushed design decision. I actually change the working directory to
// get to the corpus data. A lock needs to be held because the client can be multithreaded,
// and a race condition could cause the directory to be entered twice.
lock (DataDirectories.DirectoryControl)
{
DataDirectories.EnterDirectory(CorpusWriter.DirectoryName);
string filename = String.Format(CorpusWriter.FilenamePattern, name, order);
if (!File.Exists(filename))
throw new FileNotFoundException();
// Read each line of the data file into an individual
// Markov node.
using (var file = File.OpenText(filename))
{
while (!file.EndOfStream)
{
data.Line = file.ReadLine();
ParseLine(data);
}
}
DataDirectories.LeaveDirectory();
}
return probabilityHash;
}
