前面一节中我们说过Mapper最小输入数据单元是InputSplit。比如对于那么对于一个记录行形式的文本大于128M时,HDFS将会分成多块存储(block),同时分片并非到每行行尾。这样就会产生两个问题
1. Hadoop的一个Block默认是128M,那么对于一个记录行形式的文本,会不会造成一行记录被分到两个Block当中?
2. 在把文件从Block中读取出来进行切分时,会不会造成一行记录被分成两个InputSplit,如果被分成两个InputSplit,这样一个InputSplit里面就有一行不完整的数据,那么处理这个InputSplit的Mapper会不会得出不正确的结果?
对于上面的两个问题,首先再次明确两个概念:Block和InputSplit:
1. Block是HDFS存储文件的单位(默认是128M);
2. InputSplit是MapReduce对文件进行处理和运算的输入单位,只是一个逻辑概念,每个InputSplit并没有对文件实际的切割,只是记录了要处理的数据的位置(包括文件的path和hosts)和长度(由start和length决定)。
因此以行记录形式的文本,可能存在一行记录被划分到不同的Block,甚至不同的DataNode上去。通过分析FileInputFormat里面的getSplits方法,可以得出,某一行记录同样也可能被划分到不同的InputSplit。
我们在再从org.apache.hadoop.mapreduce.lib.input.FileInputFormat源码分析
/** * Generate the list of files and make them into FileSplits. * @param job the job context * @throws IOException */ public List<InputSplit> getSplits(JobContext job) throws IOException { long minSize = Math.max(getFormatMinSplitSize(), getMinSplitSize(job)); long maxSize = getMaxSplitSize(job); // generate splits List<InputSplit> splits = new ArrayList<InputSplit>(); List<FileStatus> files = listStatus(job); for (FileStatus file: files) { Path path = file.getPath(); long length = file.getLen(); if (length != 0) { BlockLocation[] blkLocations; if (file instanceof LocatedFileStatus) { blkLocations = ((LocatedFileStatus) file).getBlockLocations(); } else { FileSystem fs = path.getFileSystem(job.getConfiguration()); blkLocations = fs.getFileBlockLocations(file, 0, length); } if (isSplitable(job, path)) { long blockSize = file.getBlockSize(); long splitSize = computeSplitSize(blockSize, minSize, maxSize); long bytesRemaining = length; while (((double) bytesRemaining)/splitSize > SPLIT_SLOP) { int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining); splits.add(makeSplit(path, length-bytesRemaining, splitSize, blkLocations[blkIndex].getHosts())); bytesRemaining -= splitSize; } if (bytesRemaining != 0) { int blkIndex = getBlockIndex(blkLocations, length-bytesRemaining); splits.add(makeSplit(path, length-bytesRemaining, bytesRemaining, blkLocations[blkIndex].getHosts())); } } else { // not splitable splits.add(makeSplit(path, 0, length, blkLocations[0].getHosts())); } } else { //Create empty hosts array for zero length files splits.add(makeSplit(path, 0, length, new String[0])); } } // Save the number of input files for metrics/loadgen job.getConfiguration().setLong(NUM_INPUT_FILES, files.size()); LOG.debug("Total # of splits: " + splits.size()); return splits; }
从上面的代码可以看出,对文件进行切分其实很简单:获取文件在HDFS上的路径和Block信息,然后根据splitSize对文件进行切分,splitSize = computeSplitSize(blockSize, minSize, maxSize);maxSize,minSize,blockSize都可以配置,默认splitSize 就等于blockSize的默认值(128m)。
FileInputFormat对文件的切分是严格按照偏移量来的,因此一行记录比较长的话,可能被切分到不同的InputSplit。 但这并不会对Map造成影响,尽管一行记录可能被拆分到不同的InputSplit,但是与FileInputFormat关联的RecordReader被设计的足够健壮,当一行记录跨InputSplit时,其能够到读取不同的InputSplit,直到把这一行记录读取完成 。我们拿最常见的TextInputFormat源码分析如何处理跨行InputSplit的,TextInputFormat关联的是LineRecordReader,下面我们先看LineRecordReader的的nextKeyValue方法里读取文件的代码:
public boolean nextKeyValue() throws IOException { if (key == null) { key = new LongWritable(); } key.set(pos); if (value == null) { value = new Text(); } int newSize = 0; // We always read one extra line, which lies outside the upper // split limit i.e. (end - 1) while (getFilePosition() <= end) { newSize = in.readLine(value, maxLineLength, Math.max(maxBytesToConsume(pos), maxLineLength)); pos += newSize; if (newSize < maxLineLength) { break; } // line too long. try again LOG.info("Skipped line of size " + newSize + " at pos " + (pos - newSize)); } if (newSize == 0) { key = null; value = null; return false; } else { return true; } }
1、其读取文件是通过LineReader(in就是一个LineReader实例)的readLine方法完成的。关键的逻辑就在这个readLine方法里,这个方法主要的逻辑归纳起来是3点:
- 总是从buffer里读取数据,如果buffer里的数据读完了,先加载下一批数据到buffer
- 在buffer中查找"行尾",将开始位置至行尾处的数据拷贝给str(也就是最后的Value).如果为遇到"行尾",继续加载新的数据到buffer进行查找
- 关键点在于:给到buffer的数据是直接从文件中读取的,完全不会考虑是否超过了split的界限,而是一直读取到当前行结束为止
/** * Read a line terminated by one of CR, LF, or CRLF. */ private int readDefaultLine(Text str, int maxLineLength, int maxBytesToConsume) throws IOException { /* We're reading data from in, but the head of the stream may be * already buffered in buffer, so we have several cases: * 1. No newline characters are in the buffer, so we need to copy * everything and read another buffer from the stream. * 2. An unambiguously terminated line is in buffer, so we just * copy to str. * 3. Ambiguously terminated line is in buffer, i.e. buffer ends * in CR. In this case we copy everything up to CR to str, but * we also need to see what follows CR: if it's LF, then we * need consume LF as well, so next call to readLine will read * from after that. * We use a flag prevCharCR to signal if previous character was CR * and, if it happens to be at the end of the buffer, delay * consuming it until we have a chance to look at the char that * follows. */ str.clear(); int txtLength = 0; //tracks str.getLength(), as an optimization int newlineLength = 0; //length of terminating newline boolean prevCharCR = false; //true of prev char was CR long bytesConsumed = 0; do { int startPosn = bufferPosn; //starting from where we left off the last time //如果buffer中的数据读完了,先加载一批数据到buffer里 if (bufferPosn >= bufferLength) { startPosn = bufferPosn = 0; if (prevCharCR) { ++bytesConsumed; //account for CR from previous read } bufferLength = in.read(buffer); if (bufferLength <= 0) { break; // EOF } } //注意:由于不同操作系统对“行结束符“的定义不同: //UNIX: '\n' (LF) //Mac: '\r' (CR) //Windows: '\r\n' (CR)(LF) //为了准确判断一行的结尾,程序的判定逻辑是: //1.如果当前符号是LF,可以确定一定是到了行尾,但是需要参考一下前一个 //字符,因为如果前一个字符是CR,那就是windows文件,“行结束符的长度” //(即变量:newlineLength)应该是2,否则就是UNIX文件,“行结束符的长度”为1。 //2.如果当前符号不是LF,看一下前一个符号是不是CR,如果是也可以确定一定上个字符就是行尾了,这是一个mac文件。 //3.如果当前符号是CR的话,还需要根据下一个字符是不是LF判断“行结束符的长度”,所以只是标记一下prevCharCR=true,供读取下个字符时参考 for (; bufferPosn < bufferLength; ++bufferPosn) { //search for newline if (buffer[bufferPosn] == LF) {//存在'\n'换行字符 newlineLength = (prevCharCR) ? 2 : 1; ++bufferPosn; // at next invocation proceed from following byte break; } if (prevCharCR) { //CR + notLF, we are at notLF newlineLength = 1; break; } prevCharCR = (buffer[bufferPosn] == CR);//存在'\r'回车字符 } int readLength = bufferPosn - startPosn; if (prevCharCR && newlineLength == 0) { --readLength; //CR at the end of the buffer } bytesConsumed += readLength; int appendLength = readLength - newlineLength; if (appendLength > maxLineLength - txtLength) { appendLength = maxLineLength - txtLength; } if (appendLength > 0) { str.append(buffer, startPosn, appendLength); txtLength += appendLength; } //newlineLength == 0 就意味着始终没有读到行尾,程序会继续通过文件输入流继续从文件里读取数据。 //这里有一个非常重要的地方:in的实例创建自构造函数:org.apache.hadoop.mapreduce.LineRecordReader.lib.input.LineRecordReader.initialize(InputSplit, TaskAttemptContext) //方法内:FSDataInputStream fileIn = fs.open(split.getPath()); 我们以看到: //对于LineRecordReader:当它对取“一行”时,一定是读取到完整的行,不会受filesplit的任何影响,因为它读取是filesplit所在的文件,而不是限定在filesplit的界限范围内。 //所以不会出现“断行”的问题! } while (newlineLength == 0 && bytesConsumed < maxBytesToConsume); if (bytesConsumed > (long)Integer.MAX_VALUE) { throw new IOException("Too many bytes before newline: " + bytesConsumed); } return (int)bytesConsumed; }2、按照readLine的上述行为,在遇到跨split的行时,会将下一个split开始行数据读取出来构成一行完整的数据,那么下一个split怎么判定开头的一行有没有被上一个split的LineRecordReader读取过从而避免漏读或重复读取开头一行呢?这方面LineRecordReader使用了一个简单而巧妙的方法:既然无法断定每一个split开始的一行是独立的一行还是被切断的一行的一部分,那就跳过每个split的开始一行(当然要除第一个split之外),从第二行开始读取,然后在到达split的结尾端时总是再多读一行,这样数据既能接续起来又避开了断行带来的麻烦.以下是相关的源码:
// If this is not the first split, we always throw away first record // because we always (except the last split) read one extra line in // next() method. if (start != 0) {//非第一个InputSplit忽略掉第一行 start += in.readLine(new Text(), 0, maxBytesToConsume(start)); } this.pos = start;3、相应地,在LineRecordReader判断是否还有下一行的方法:org.apache.hadoop.mapreduce.lib.input.LineRecordReader.nextKeyValue()中,while使用的判定条件保证了InputSplit读取跨界的问题:当前位置小于或等于split的结尾位置,也就说:当前已处于split的结尾位置上时,while依然会执行一次,这一次读到显然已经是下一个split的开始行了。
public boolean nextKeyValue() throws IOException { if (key == null) { key = new LongWritable(); } key.set(pos); if (value == null) { value = new Text(); } int newSize = 0; // We always read one extra line, which lies outside the upper // split limit i.e. (end - 1) while (getFilePosition() <= end) {//保证InputSplit读取边界的问题 newSize = in.readLine(value, maxLineLength, Math.max(maxBytesToConsume(pos), maxLineLength)); pos += newSize; if (newSize < maxLineLength) { break; } // line too long. try again LOG.info("Skipped line of size " + newSize + " at pos " + (pos - newSize)); } if (newSize == 0) { key = null; value = null; return false; } else { return true; } }至此,通过上面的源码分析我们清楚了解到 TextInputFormat是如何解决跨行Block和InputSplit的,因此当我们需要实现自己的InputFormat时,也会面临在切分数据时的连续性解析问题。