public void Emotions()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Add<EmotionToken>();
EmotionToken[] tokens = tokenizer.Tokenize("Šodien esmu :) bet vakar biju :(").OfType<EmotionToken>().ToArray();
Assert.AreEqual(":)", tokens[0].Text);
Assert.AreEqual(":(", tokens[1].Text);
}
public void EqualMatch()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Clear();
tokenizer.Add<TimeSpanToken>(); // matches 00:00:00
tokenizer.Add<ClockToken>(); // matches 00:00:00
Token token = tokenizer.Tokenize("00:00:00").First();
Assert.IsTrue(token is TimeSpanToken);
}
public void kā_arī()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Add<LuMiiCollocationToken>();
tokenizer.Compile();
string text = "Es skrienu kā arī lecu.";
Token[] tokens = tokenizer.Tokenize(text).ToArray();
Assert.AreEqual(5, tokens.Length);
Assert.AreEqual("kā arī", tokens[2].Text);
}
public void ļcien()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Add<LuMiiAbbreviationToken>();
tokenizer.Compile();
string text = "Ļ.cien. kungs!";
Token[] tokens = tokenizer.Tokenize(text).ToArray();
Assert.AreEqual(3, tokens.Length);
Assert.AreEqual("Ļ.cien.", tokens[0].Text);
}
public void Compile()
{
int count = 100;
Stopwatch timer = new Stopwatch();
timer.Start();
for (int i = 0; i < count; i++)
{
LatvianTokenizer tokenizer = new LatvianTokenizer();
}
timer.Stop();
Debug.WriteLine("Compile: {0:0.000} ms", timer.ElapsedMilliseconds / count);
Debug.WriteLine("Compile x{1}: {0:0.000} ms", timer.ElapsedMilliseconds, count);
}
public void kaut_jel_milj_kgs()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Add<LuMiiAbbreviationToken>();
tokenizer.Add<LuMiiCollocationToken>();
tokenizer.Compile();
string text = "kaut JEL 2 milj. mans k-gs";
Token[] tokens = tokenizer.Tokenize(text).ToArray();
Assert.AreEqual(5, tokens.Length);
Assert.AreEqual("kaut JEL", tokens[0].Text);
Assert.AreEqual("milj.", tokens[2].Text);
Assert.AreEqual("k-gs", tokens[4].Text);
}
public void Load()
{
string filename = Path.GetTempFileName();
LatvianTokenizer tokenizer = new LatvianTokenizer();
tokenizer.Save(filename);
int count = 100;
Stopwatch timer = new Stopwatch();
timer.Start();
for (int i = 0; i < count; i++)
{
LatvianTokenizer tokenizer2 = new LatvianTokenizer(filename);
}
timer.Stop();
Debug.WriteLine("Load: {0:0.000} ms", timer.ElapsedMilliseconds / count);
Debug.WriteLine("Load x{1}: {0:0.000} ms", timer.ElapsedMilliseconds, count);
File.Delete(filename);
}
public void InitialsRemoved()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Remove<InitialsToken>();
tokenizer.Compile(); // optional
Token[] tokens = tokenizer.Tokenize("A.Bērziņš").ToArray();
Assert.AreEqual("A", tokens[0].Text);
Assert.AreEqual(".", tokens[1].Text);
Assert.AreEqual("Bērziņš", tokens[2].Text);
}
public void Initials()
{
Token[] tokens = new LatvianTokenizer().Tokenize("A.Bērziņš").ToArray();
Assert.AreEqual("A.", tokens[0].Text);
Assert.AreEqual("Bērziņš", tokens[1].Text);
}
public void Position()
{
Token[] tokens = new LatvianTokenizer().Tokenize("Vārds.").ToArray();
Assert.AreEqual(0, tokens[0].Position);
Assert.AreEqual(5, tokens[0].PositionEnd);
Assert.AreEqual(0, tokens[0].Line);
Assert.AreEqual(0, tokens[0].LineEnd);
Assert.AreEqual(0, tokens[0].LinePosition);
Assert.AreEqual(5, tokens[0].LinePositionEnd);
Assert.AreEqual(".", tokens[1].Text);
Assert.AreEqual(5, tokens[1].Position);
Assert.AreEqual(6, tokens[1].PositionEnd);
Assert.AreEqual(0, tokens[1].Line);
Assert.AreEqual(0, tokens[1].LineEnd);
Assert.AreEqual(5, tokens[1].LinePosition);
Assert.AreEqual(6, tokens[1].LinePositionEnd);
}
public void WhitespaceIncluded()
{
LatvianTokenizer tokenizer = new LatvianTokenizer() { IncludeWhitespace = true };
Token[] tokens = tokenizer.Tokenize("123 456").ToArray();
Assert.AreEqual("123", tokens[0].Text);
Assert.AreEqual(" ", tokens[1].Text);
Assert.AreEqual("456", tokens[2].Text);
}
public void QuickStart()
{
List<Token> tokens = new List<Token>();
DateToken dateToken = null;
Debug.WriteLine("Latvian.Tests.Tokenization.ReadmeTests.QuickStart starts");
string text = "Sveika, pasaule! Man iet labi. Šodienas datums ir 2014-01-01";
LatvianTokenizer tokenizer = new LatvianTokenizer();
foreach (Token token in tokenizer.Tokenize(text))
{
Debug.WriteLine("Line {0}: Pos {1}: Type: {2} Token: {3}",
token.Line, token.Position, token.GetType(), token.Text);
tokens.Add(token);
if (token is DateToken)
{
dateToken = token as DateToken;
Debug.WriteLine(dateToken.DateTime.ToString("dd/MM/yyyy"));
}
}
Debug.WriteLine("Latvian.Tests.Tokenization.ReadmeTests.QuickStart end");
Assert.AreEqual(12, tokens.Count);
Assert.AreEqual("Sveika", tokens[0].Text);
Assert.AreEqual(typeof(WordToken), tokens[0].GetType());
Assert.AreEqual(0, tokens[0].Position);
Assert.AreEqual(6, tokens[0].PositionEnd);
Assert.AreEqual(0, tokens[0].Line);
Assert.AreEqual(0, tokens[0].LineEnd);
Assert.AreEqual(0, tokens[0].LinePosition);
Assert.AreEqual(6, tokens[0].LinePositionEnd);
Assert.AreEqual(",", tokens[1].Text);
Assert.AreEqual(typeof(PunctuationToken), tokens[1].GetType());
Assert.AreEqual(6, tokens[1].Position);
Assert.AreEqual(7, tokens[1].PositionEnd);
Assert.AreEqual(0, tokens[1].Line);
Assert.AreEqual(0, tokens[1].LineEnd);
Assert.AreEqual(6, tokens[1].LinePosition);
Assert.AreEqual(7, tokens[1].LinePositionEnd);
Assert.AreEqual("pasaule", tokens[2].Text);
Assert.AreEqual(typeof(WordToken), tokens[2].GetType());
Assert.AreEqual(8, tokens[2].Position);
Assert.AreEqual(15, tokens[2].PositionEnd);
Assert.AreEqual(0, tokens[2].Line);
Assert.AreEqual(0, tokens[2].LineEnd);
Assert.AreEqual(8, tokens[2].LinePosition);
Assert.AreEqual(15, tokens[2].LinePositionEnd);
Assert.AreEqual("!", tokens[3].Text);
Assert.AreEqual(typeof(PunctuationToken), tokens[3].GetType());
Assert.AreEqual(15, tokens[3].Position);
Assert.AreEqual(16, tokens[3].PositionEnd);
Assert.AreEqual(0, tokens[3].Line);
Assert.AreEqual(0, tokens[3].LineEnd);
Assert.AreEqual(15, tokens[3].LinePosition);
Assert.AreEqual(16, tokens[3].LinePositionEnd);
Assert.AreEqual("Man", tokens[4].Text);
Assert.AreEqual(typeof(WordToken), tokens[4].GetType());
Assert.AreEqual(17, tokens[4].Position);
Assert.AreEqual(20, tokens[4].PositionEnd);
Assert.AreEqual(0, tokens[4].Line);
Assert.AreEqual(0, tokens[4].LineEnd);
Assert.AreEqual(17, tokens[4].LinePosition);
Assert.AreEqual(20, tokens[4].LinePositionEnd);
Assert.AreEqual("iet", tokens[5].Text);
Assert.AreEqual(typeof(WordToken), tokens[5].GetType());
Assert.AreEqual(21, tokens[5].Position);
Assert.AreEqual(24, tokens[5].PositionEnd);
Assert.AreEqual(0, tokens[5].Line);
Assert.AreEqual(0, tokens[5].LineEnd);
Assert.AreEqual(21, tokens[5].LinePosition);
Assert.AreEqual(24, tokens[5].LinePositionEnd);
Assert.AreEqual("labi", tokens[6].Text);
Assert.AreEqual(typeof(WordToken), tokens[6].GetType());
Assert.AreEqual(25, tokens[6].Position);
Assert.AreEqual(29, tokens[6].PositionEnd);
Assert.AreEqual(0, tokens[6].Line);
Assert.AreEqual(0, tokens[6].LineEnd);
Assert.AreEqual(25, tokens[6].LinePosition);
Assert.AreEqual(29, tokens[6].LinePositionEnd);
Assert.AreEqual(".", tokens[7].Text);
Assert.AreEqual(typeof(PunctuationToken), tokens[7].GetType());
Assert.AreEqual(29, tokens[7].Position);
Assert.AreEqual(30, tokens[7].PositionEnd);
Assert.AreEqual(0, tokens[7].Line);
Assert.AreEqual(0, tokens[7].LineEnd);
Assert.AreEqual(29, tokens[7].LinePosition);
Assert.AreEqual(30, tokens[7].LinePositionEnd);
Assert.AreEqual("2014-01-01", tokens[11].Text);
Assert.AreEqual(typeof(DateToken), tokens[11].GetType());
Assert.AreEqual(50, tokens[11].Position);
Assert.AreEqual(60, tokens[11].PositionEnd);
Assert.AreEqual(0, tokens[11].Line);
Assert.AreEqual(0, tokens[11].LineEnd);
Assert.AreEqual(50, tokens[11].LinePosition);
Assert.AreEqual(60, tokens[11].LinePositionEnd);
Assert.AreEqual(tokens[11], dateToken);
Assert.AreEqual("01.01.2014", dateToken.DateTime.ToString("dd/MM/yyyy"));
}
public void QuickStart_TokenizeSentences()
{
string text = "Sveika, pasaule! Man iet labi. Šodienas datums ir 2014-01-01";
List<Sentence> sentences = new List<Sentence>();
LatvianTokenizer tokenizer = new LatvianTokenizer();
foreach (Sentence sentence in tokenizer.TokenizeSentences(text))
{
List<Token> sentenceTokens = new List<Token>();
foreach (Token token in sentence)
{
sentenceTokens.Add(token);
}
sentences.Add(sentenceTokens);
}
Assert.AreEqual(3, sentences.Count());
Assert.AreEqual(4, sentences[0].Count());
Assert.AreEqual(4, sentences[1].Count());
Assert.AreEqual(4, sentences[2].Count());
}
public void LongestMatch()
{
Token[] tokens = new LatvianTokenizer().Tokenize("2014-01-01 2014-01-01T12:00:00").ToArray();
Assert.AreEqual("2014-01-01", tokens[0].Text);
Assert.AreEqual("2014-01-01T12:00:00", tokens[1].Text);
}
public void Stream()
{
using (Stream stream = new MemoryStream())
{
using (StreamWriter writer = new StreamWriter(stream, Encoding.UTF8, bufferSize: 1024, leaveOpen: true))
writer.Write("123 456");
stream.Position = 0;
Token[] tokens = new LatvianTokenizer().Tokenize(stream).ToArray();
Assert.AreEqual("123", tokens[0].Text);
Assert.AreEqual("456", tokens[1].Text);
}
}
public void TokenizeString()
{
int count = 5;
int mebibytes = 10;
int size = mebibytes * 1024 * 1024;
string text = DarbaLikums(size);
Assert.AreEqual(size, text.Length);
LatvianTokenizer tokenizer = new LatvianTokenizer();
Stopwatch timer = new Stopwatch();
int tokenCount = 0;
timer.Start();
for (int i = 0; i < count; i++)
{
foreach (Token token in tokenizer.Tokenize(text))
{
tokenCount++;
}
}
timer.Stop();
Assert.IsTrue(tokenCount > 0);
Debug.WriteLine("Tokenize string ({1} MiB): {0:0.000} ms", timer.ElapsedMilliseconds / count, mebibytes);
Debug.WriteLine("Tokenize string: {0:0.000} MB/s", (mebibytes * count) / timer.Elapsed.TotalSeconds, mebibytes);
Debug.WriteLine("Tokenize string: {0:0.000} tokens/s", tokenCount / timer.Elapsed.TotalSeconds);
}
public void Distinct()
{
var tokens = new LatvianTokenizer().Tokenize("viens viens").Distinct();
Assert.AreEqual(1, tokens.Count());
}
public void Reorder()
{
LatvianTokenizer tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Clear();
tokenizer.Add<TimeSpanToken>(); // matches 00:00:00
tokenizer.Add<ClockToken>(); // matches 00:00:00
tokenizer.Remove<ClockToken>();
tokenizer.Insert<ClockToken>(0);
Token token = tokenizer.Tokenize("00:00:00").First();
Assert.IsTrue(token is ClockToken);
tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Clear();
tokenizer.Add<TimeSpanToken>(); // matches 00:00:00
tokenizer.Add<ClockToken>(); // matches 00:00:00
tokenizer.Remove(typeof(ClockToken));
tokenizer.Insert(0, typeof(ClockToken));
token = tokenizer.Tokenize("00:00:00").First();
Assert.IsTrue(token is ClockToken);
tokenizer = new LatvianTokenizer(compile: false);
tokenizer.Clear();
tokenizer.Add<TimeSpanToken>(); // matches 00:00:00
tokenizer.Add<ClockToken>(); // matches 00:00:00
tokenizer.Move<ClockToken>(0);
token = tokenizer.Tokenize("00:00:00").First();
Assert.IsTrue(token is ClockToken);
}
public void DateToken()
{
Token token = new LatvianTokenizer().Tokenize("2014-01-01").First();
Assert.IsTrue(token is DateToken);
Assert.AreEqual(2014, ((DateToken)token).DateTime.Year);
}
public void OrdinalNumberToken()
{
Token token = new LatvianTokenizer().Tokenize("123").First();
Assert.IsTrue(token is OrdinalNumberToken);
Assert.AreEqual(123 * 2, ((OrdinalNumberToken)token).Value * 2);
}
public void WhitespaceExcluded()
{
Token[] tokens = new LatvianTokenizer().Tokenize("123 456").ToArray();
Assert.AreEqual("123", tokens[0].Text);
Assert.AreEqual("456", tokens[1].Text);
}
public void Text()
{
Token token = new LatvianTokenizer().Tokenize("vārds").First();
Assert.AreEqual("vārds", token.Text);
Assert.AreEqual("vārds", token.ToString());
}
public void QuickStart_BreakSentences()
{
string text = "Sveika, pasaule! Man iet labi. Šodienas datums ir 2014-01-01";
LatvianTokenizer tokenizer = new LatvianTokenizer();
Token[] tokens = tokenizer.Tokenize(text).ToArray();
Sentence[] sentences = tokenizer.BreakSentences(tokens).ToArray();
Assert.AreEqual(3, sentences.Count());
Assert.AreEqual(4, sentences[0].Count());
Assert.AreEqual(4, sentences[1].Count());
Assert.AreEqual(4, sentences[2].Count());
}
public void TextReader()
{
using (StringReader reader = new StringReader("123 456"))
{
Token[] tokens = new LatvianTokenizer().Tokenize(reader).ToArray();
Assert.AreEqual("123", tokens[0].Text);
Assert.AreEqual("456", tokens[1].Text);
}
}
public void Values()
{
string[] tokens = new LatvianTokenizer().Tokenize("viens divi").Select(t => t.Text).ToArray();
Assert.AreEqual("viens", tokens[0]);
Assert.AreEqual("divi", tokens[1]);
}
public void TokenizeFile()
{
int count = 5;
int mebibytes = 10;
int size = mebibytes * 1024 * 1024; ;
string text = DarbaLikums(size);
Assert.AreEqual(size, text.Length);
string filename = Path.GetTempFileName();
File.WriteAllText(filename, text);
Assert.AreEqual(size, File.ReadAllText(filename).Length);
LatvianTokenizer tokenizer = new LatvianTokenizer();
Stopwatch timer = new Stopwatch();
int tokenCount = 0;
timer.Start();
for (int i = 0; i < count; i++)
{
using (StreamReader reader = new StreamReader(filename))
{
foreach (Token token in tokenizer.Tokenize(reader))
{
tokenCount++;
}
}
}
timer.Stop();
Assert.IsTrue(tokenCount > 0);
Debug.WriteLine("Tokenize file ({1} MiB): {0:0.000} ms", timer.ElapsedMilliseconds / count, mebibytes);
Debug.WriteLine("Tokenize file: {0:0.000} MB/s", (mebibytes * count) / timer.Elapsed.TotalSeconds, mebibytes);
Debug.WriteLine("Tokenize file: {0:0.000} tokens/s", tokenCount / timer.Elapsed.TotalSeconds);
File.Delete(filename);
}
public void LoadSave()
{
string filename = Path.GetTempFileName();
LatvianTokenizer tokenizer = new LatvianTokenizer();
tokenizer.Save(filename);
tokenizer = new LatvianTokenizer(filename);
Token[] tokens = tokenizer.Tokenize("123 456").ToArray();
Assert.AreEqual("123", tokens[0].Text);
Assert.AreEqual("456", tokens[1].Text);
}