Deal correctly with zero length sequences in per tile quality.

Relates to s-andrews#57

Author: Andrewss

uk/ac/babraham/FastQC/Modules/PerTileQualityScores.java

@@ -146,21 +146,27 @@ private synchronized void getPercentages () {
 		char minChar = 0;
 		char maxChar = 0;
 
-		// Use the data from the first tile
-		QualityCount [] qualityCounts = perTileQualityCounts.get(perTileQualityCounts.keySet().toArray()[0]);
-
-		for (int q=0;q<qualityCounts.length;q++) {
-			if (q == 0) {
-				minChar = qualityCounts[q].getMinChar();
-				maxChar = qualityCounts[q].getMaxChar();
-			}
-			else {
-				if (qualityCounts[q].getMinChar() < minChar) {
+		// Iterate through the tiles to check them all in case
+		// we're dealing with unrepresentative data in the first one.
+		Iterator<QualityCount[]> qit = perTileQualityCounts.values().iterator();
+		
+		while (qit.hasNext()) { 
+			
+			QualityCount [] qualityCounts = qit.next();
+	
+			for (int q=0;q<qualityCounts.length;q++) {
+				if (minChar == 0) {
 					minChar = qualityCounts[q].getMinChar();
-				}
-				if (qualityCounts[q].getMaxChar() > maxChar) {
 					maxChar = qualityCounts[q].getMaxChar();
 				}
+				else {
+					if (qualityCounts[q].getMinChar() < minChar) {
+						minChar = qualityCounts[q].getMinChar();
+					}
+					if (qualityCounts[q].getMaxChar() > maxChar) {
+						maxChar = qualityCounts[q].getMaxChar();
+					}
+				}
 			}
 		}
 
@@ -179,6 +185,12 @@ public void processSequence(Sequence sequence) {
 
 		// Don't waste time calculating this if we're not going to use it anyway
 		if (ignoreInReport) return;
+		
+		// Don't bother with sequences with zero length as they don't have any 
+		// quality information anyway.
+		if (sequence.getQualityString().length() == 0) {
+			return;
+		}
 
 		calculated = false;
 
@@ -192,10 +204,12 @@ public void processSequence(Sequence sequence) {
 
 		// This module does quite a lot of work and ends up being the limiting
 		// step when calculating.  We'll therefore take only a sample of the 
-		// sequences to try to get a representative selection.
+		// sequences to try to get a representative selection.  We'll use the 
+		// first 10k sequences in case we're dealing with a very small file
+		// and then take 10% of the rest.
 
 		++totalCount;
-		if (totalCount % 10 != 0) return;
+		if (totalCount > 10000 && totalCount % 10 != 0) return;
 
 		// First try to split the id by :
 		int tile = 0;

uk/ac/babraham/FastQC/Sequence/QualityEncoding/PhredEncoding.java

@@ -45,7 +45,7 @@ else if (lowestChar == ILLUMINA_1_3_ENCODING_OFFSET+1) {
 		else if (lowestChar <= 126) {
 			return new PhredEncoding("Illumina 1.5", ILLUMINA_1_3_ENCODING_OFFSET);
 		}
-		throw new IllegalArgumentException("No known encodings with chars > 126 (Yours was "+lowestChar+")");
+		throw new IllegalArgumentException("No known encodings with chars > 126 (Yours was "+lowestChar+" with value "+(int)lowestChar+")");
 	}
 
 	public static double convertSangerPhredToProbability (int phred) {

uk/ac/babraham/FastQC/Utilities/QualityCount.java

@@ -25,7 +25,7 @@ public class QualityCount {
 	/*
 	 * So I'm on my third go at writing this.  I've now tried an all
 	 * primitive version of this class so that we don't have to do 
-	 * hash lookps which require a conversion from chr to Character.
+	 * hash lookups which require a conversion from chr to Character.
 	 * We should also be safe with 150 slots which will give us up to
 	 * Phred 86 with a 64 offset, which should be plenty.
 	 */