Member joshua.decoder.ff.lm.AbstractLM.logProbOfBackoffState (List< Integer > ngram, int order, int qtyAdditionalBackoffWeight): this function is much slower than the int[] version
Member joshua.decoder.ff.lm.DefaultNGramLanguageModel.ngramLogProbability (List< Integer > ngram, int order): this function is much slower than the int[] version
Member joshua.decoder.ff.lm.kenlm.jni.KenLM.ngramLogProbability (List< Integer > ngram, int order): pass int arrays to prob instead.
Member joshua.decoder.ff.lm.kenlm.jni.KenLM.sentenceLogProbability (List< Integer > sentence, int order, int startIndex): pass int arrays to prob instead.
Member joshua.util.FileUtility.getReadFileStream (String filename): use joshua.util.io.LineReader instead.
Member joshua.util.FileUtility.read_line_lzf (BufferedReader in): use joshua.util.io.LineReader instead.
Member joshua.util.io.LineReader.getInputStream (String filename): This method is provided in order for joshua.decoder.DecoderThread to open files in the canonical way for handing off to joshua.decoder.segment_file.SegmentFileParser. The SegmentFileParser interface can't be made more liberal (e.g. to accept a java.io.Reader) because javax.xml.parsers.SAXParser can't parse that argument and no common java.io.Reader gives access to the underlying InputStream. This method is considered a hack which should be removed once a better solution presents itself.