private WordOccurenceCollection CountOccurencesForText(string textBlockText)
{
if (textBlockText == null)
{
throw new ArgumentNullException(nameof(textBlockText));
}
var occurences = new WordOccurenceCollection();
// Split text to words
var words =
textBlockText.Split(new[] { '.', '?', '!', ' ', ':', ',' }, StringSplitOptions.RemoveEmptyEntries)
.Where(x => x.Length >= MinimumWordLength);
foreach (var word in words)
{
var simpleWord = word.Simplify();
if (!string.IsNullOrWhiteSpace(simpleWord) && simpleWord.Length >= MinimumWordLength)
{
occurences.Add(simpleWord);
}
}
return(occurences);
}
public void GetWordCount_OnExecuteWithNull_ThrowsException()
{
var wordOccurendeCollection = new WordOccurenceCollection {
"word"
};
wordOccurendeCollection.GetWordCount(null);
}
public void GetWordCount_OnExecuteWithWordNotInCollection_ReturnsZero()
{
var wordOccurendeCollection = new WordOccurenceCollection {
"word"
};
var result = wordOccurendeCollection.GetWordCount("test");
Assert.AreEqual(0, result);
}
public void Merge_OnExecute_ReturnsNewMergedCollection()
{
var wordOccurendeCollection = new WordOccurenceCollection {
"word", "word"
};
var wordOccurendeCollection2 = new WordOccurenceCollection {
"word", "word"
};
var mergedCollection = wordOccurendeCollection.Merge(wordOccurendeCollection2);
Assert.AreNotEqual(wordOccurendeCollection, mergedCollection);
Assert.AreNotEqual(wordOccurendeCollection2, mergedCollection);
Assert.AreEqual(4, mergedCollection.GetWordCount("word"));
}
public IEnumerable <KeyValuePair <string, int> > GetKeywords(HtmlResult html)
{
var occurences = new WordOccurenceCollection();
var textBlocks = html.Document.SelectNodes("//*[not(self::script) and not(self::style)]]//text()");
if (textBlocks != null)
{
foreach (var textBlock in textBlocks)
{
var textBlockText = textBlock.InnerText;
var occurencesInBlock = CountOccurencesForText(textBlockText);
occurences = occurences.Merge(occurencesInBlock);
}
}
return(occurences.OrderByDescending(x => x.Value));
}
public void GetWordCount_OnExecute_ReturnsTheOccurenceOfTheWords()
{
var wordOccurendeCollection = new WordOccurenceCollection
{
"word",
"word",
"word",
"word",
"test",
"test",
"test"
};
var resultForWord = wordOccurendeCollection.GetWordCount("word");
var resultForTest = wordOccurendeCollection.GetWordCount("test");
Assert.AreEqual(4, resultForWord);
Assert.AreEqual(3, resultForTest);
}
public WordOccurenceCollection GetKeywords(HtmlNode htmlNode)
{
if (htmlNode == null)
{
throw new ArgumentNullException(nameof(htmlNode));
}
var occurences = new WordOccurenceCollection();
var textBlocks = htmlNode.SelectNodes("//*[not(self::script) and not(self::style)]//text()");
if (textBlocks != null)
{
var textBlocksWithText = textBlocks.Where(x => !string.IsNullOrWhiteSpace(x.InnerText)).Select(x => x.InnerHtml);
foreach (var text in textBlocksWithText)
{
var occurencesInBlock = CountOccurencesForText(text);
occurences = occurences.Merge(occurencesInBlock);
}
}
return(occurences);
}
