Mercurial > hg > monetdb-java

changeset 204:ae1d0d1c2f0f

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Optimize TupleLineParser by doing less copying of string data when field value does not contain an escape character, which is the case for most strings.
author Martin van Dinther <martin.van.dinther@monetdbsolutions.com>
date Thu, 01 Mar 2018 18:44:06 +0100 (2018-03-01)
parents c6abd650aeb4
children 5b13ccaba741 94131372ad0b
files src/main/java/nl/cwi/monetdb/mcl/parser/TupleLineParser.java
diffstat 1 files changed, 106 insertions(+), 81 deletions(-) [+]
line wrap: on
line diff
--- a/src/main/java/nl/cwi/monetdb/mcl/parser/TupleLineParser.java
+++ b/src/main/java/nl/cwi/monetdb/mcl/parser/TupleLineParser.java
@@ -38,13 +38,10 @@ public class TupleLineParser extends MCL
 	 */
 	@Override
 	public int parse(String source) throws MCLParseException {
-		int len = source.length();
-		char[] chrLine = new char[len];
-		source.getChars(0, len, chrLine, 0);
-
+		final int len = source.length();
 		// first detect whether this is a single value line (=) or a
 		// real tuple ([)
-		if (chrLine[0] == '=') {
+		if (len >= 1 && source.charAt(0) == '=') {
 			if (values.length != 1)
 				throw new MCLParseException(values.length +
 						" columns expected, but only single value found");
@@ -54,21 +51,22 @@ public class TupleLineParser extends MCL
 
 			// reset colnr
 			reset();
-
 			return 0;
 		}
 
-		// extract separate fields by examining string, char for char
-		boolean inString = false, escaped = false;
-		int cursor = 2, column = 0;
-		StringBuilder uesc = new StringBuilder();
+		if (!source.startsWith("["))
+			throw new MCLParseException("Expected a data row starting with [");
+
+		// It is a tuple. Extract separate fields by examining the string data char for char
+		final char[] chrLine = source.toCharArray();	// convert whole string to char[] to avoid overhead of source.charAt(i) calls TODO: measure the overhead
+		boolean inString = false, escaped = false, fieldHasEscape = false;
+		final StringBuilder uesc = new StringBuilder(128);	// used for building field string value when an escape is present in the field value
+		int column = 0, cursor = 2;
 		for (int i = 2; i < len; i++) {
 			switch(chrLine[i]) {
-				default:
-					escaped = false;
-					break;
 				case '\\':
 					escaped = !escaped;
+					fieldHasEscape = true;
 					break;
 				case '"':
 					/**
@@ -89,98 +87,125 @@ public class TupleLineParser extends MCL
 					} else if (!escaped) {
 						inString = false;
 					}
-
 					// reset escaped flag
 					escaped = false;
 					break;
-				case '\t':
+				case '\t':		// potential field separator found
 					if (!inString &&
-						(i > 0 && chrLine[i - 1] == ',') ||
-						(i + 1 == len - 1 && chrLine[++i] == ']')) // dirty
+						((chrLine[i - 1] == ',') ||		// found field separator: ,\t
+						 ((i + 1 == len - 1) && chrLine[++i] == ']'))) // found last field: \t]
 					{
-						// split!
+						// extract the field value as a string, without the potential escape codes
+						final int endpos = i - 2;	// minus the tab and the comma or ]
 						if (chrLine[cursor] == '"' &&
-							chrLine[i - 2] == '"')
+						    chrLine[endpos] == '"')	// field is surrounded by double quotes, so a string with possible escape codes
 						{
-							// reuse the StringBuilder by cleaning it
-							uesc.delete(0, uesc.length());
-							// prevent capacity increasements
-							uesc.ensureCapacity((i - 2) - (cursor + 1));
-							for (int pos = cursor + 1; pos < i - 2; pos++) {
-								if (chrLine[pos] == '\\' && pos + 1 < i - 2) {
-									pos++;
-									// escapedStr and GDKstrFromStr in gdk_atoms.c only
-									// support \\ \f \n \r \t \" and \377
-									switch (chrLine[pos]) {
-										case '\\':
-											uesc.append('\\');
-											break;
-										case 'f':
-											uesc.append('\f');
-											break;
-										case 'n':
-											uesc.append('\n');
-											break;
-										case 'r':
-											uesc.append('\r');
-											break;
-										case 't':
-											uesc.append('\t');
-											break;
-										case '"':
-											uesc.append('"');
-											break;
-										case '0': case '1': case '2': case '3':
-											// this could be an octal number, let's check it out
-											if (pos + 2 < i - 2 &&
-												chrLine[pos + 1] >= '0' && chrLine[pos + 1] <= '7' &&
-												chrLine[pos + 2] >= '0' && chrLine[pos + 2] <= '7'
-											) {
-												// we got an octal number
-												try {
-													uesc.append((char)(Integer.parseInt("" + chrLine[pos] + chrLine[pos + 1] + chrLine[pos + 2], 8)));
-													pos += 2;
-												} catch (NumberFormatException e) {
-													// hmmm, this point should never be reached actually...
-													throw new AssertionError("Flow error, should never try to parse non-number");
+							if (fieldHasEscape) {
+								// reuse the StringBuilder by cleaning it
+								uesc.delete(0, uesc.length());
+								// prevent capacity increasements
+								uesc.ensureCapacity(endpos - (cursor + 1));
+								// parse the field value (excluding the double quotes) and convert it to a string without any escape characters
+								for (int pos = cursor + 1; pos < endpos; pos++) {
+									char chr = chrLine[pos];
+									if (chr == '\\' && pos + 1 < endpos) {
+										// we detected an escape
+										// escapedStr and GDKstrFromStr in gdk_atoms.c only
+										// support \\ \f \n \r \t \" and \377
+										pos++;
+										chr = chrLine[pos];
+										switch (chr) {
+											case '\\':
+												uesc.append('\\');
+												break;
+											case 'f':
+												uesc.append('\f');
+												break;
+											case 'n':
+												uesc.append('\n');
+												break;
+											case 'r':
+												uesc.append('\r');
+												break;
+											case 't':
+												uesc.append('\t');
+												break;
+											case '"':
+												uesc.append('"');
+												break;
+											case '0': case '1': case '2': case '3':
+												// this could be an octal number, let's check it out
+												if (pos + 2 < endpos) {
+													char chr2 = chrLine[pos + 1];
+													char chr3 = chrLine[pos + 2];
+													if (chr2 >= '0' && chr2 <= '7' && chr3 >= '0' && chr3 <= '7') {
+														// we got an octal number between \000 and \377
+														try {
+															uesc.append((char)(Integer.parseInt("" + chr + chr2 + chr3, 8)));
+															pos += 2;
+														} catch (NumberFormatException e) {
+															// hmmm, this point should never be reached actually...
+															throw new AssertionError("Flow error, should never try to parse non-number");
+														}
+													} else {
+														// do default action if number seems not to be an octal number
+														uesc.append(chr);
+													}
+												} else {
+													// do default action if number seems not to be an octal number
+													uesc.append(chr);
 												}
-											} else {
-												// do default action if number seems not to be correct
-												uesc.append(chrLine[pos]);
-											}
-											break;
-										default:
-											// this is wrong usage of escape, just ignore the \-escape and print the char
-											uesc.append(chrLine[pos]);
-											break;
+												break;
+											default:
+												// this is wrong usage of escape, just ignore the \-escape and print the char
+												uesc.append(chr);
+												break;
+										}
+									} else {
+										uesc.append(chr);
 									}
-								} else {
-									uesc.append(chrLine[pos]);
 								}
+								// put the unescaped string in the right place
+								values[column] = uesc.toString();
+							} else {
+								// the field is a string surrounded by double quotes and without escape chars
+								cursor++;
+								String fieldVal = new String(chrLine, cursor, endpos - cursor);
+								// if (fieldVal.contains("\\")) {
+								//	throw new MCLParseException("Invalid parsing: detected a \\ in double quoted string: " + fieldVal);
+								// }
+								values[column] = fieldVal;
 							}
-
-							// put the unescaped string in the right place
-							values[column++] = uesc.toString();
-						} else if ((i - 1) - cursor == 4 && source.indexOf("NULL", cursor) == cursor) {
-							values[column++] = null;
+						} else if (((i - 1 - cursor) == 4) && source.indexOf("NULL", cursor) == cursor) {
+							// the field contains NULL, so no value
+							values[column] = null;
 						} else {
-							values[column++] = source.substring(cursor, i - 1);
+							// the field is a string NOT surrounded by double quotes and thus without escape chars
+							String fieldVal = new String(chrLine, cursor, i - 1 - cursor);
+							// if (fieldVal.contains("\\")) {
+							//	throw new MCLParseException("Invalid parsing: detected a \\ in unquoted string: " + fieldVal);
+							// }
+							values[column] = fieldVal;
 						}
 						cursor = i + 1;
+						fieldHasEscape = false;		// reset for next field scan
+						column++;
 					}
-
 					// reset escaped flag
 					escaped = false;
 					break;
-			}
-		}
+				default:
+					escaped = false;
+					break;
+			} // end of switch()
+		} // end of for()
+
 		// check if this result is of the size we expected it to be
 		if (column != values.length)
 			throw new MCLParseException("illegal result length: " + column + "\nlast read: " + (column > 0 ? values[column - 1] : "<none>"));
 
 		// reset colnr
 		reset();
-
 		return 0;
 	}
 }