Javascript Node.js EBADF使用可写流写入文件时出错

我尝试使用Node.js处理500MB的Apache日志文件，将其语法从

ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26

到

，然后写入另一个文本文件

为了更好的内存控制和性能，我使用了

fs.createReadStream

和

fs.createWriteStream

，但只将第一行写入

output.txt

，因为脚本以错误结尾：

{[Error:EBADF，write]errno:9，code:'EBADF'}

这里我发布了一些可能有助于调试的信息

input.txt的头部：

ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:18 -0600] "GET /path/of/access/ HTTP/1.1" 302 26

ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26

output.txt的内容
：

ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26
ip.ip.ip.ip - - [02/Aug/2012:05:01:18 -0600] "GET /path/of/access/ HTTP/1.1" 302 26

ip.ip.ip.ip - - [02/Aug/2012:05:01:17 -0600] "GET /path/of/access/ HTTP/1.1" 302 26

整个剧本：

var fs = require('fs');
var data ='';
var n=0;                    //For line control
var r = fs.createReadStream('./input.txt',{
    encoding: 'ascii',
    start:0,
    // end: 100000,
});
var w = fs.createWriteStream('./output.txt',{
    encoding:'ascii'
});
function put(line){         //write into w;
    ++n;
    w.write(line+'\n');
}
function end(){
    r.destroy();
    w.destroy();
}
function onData(chunk){
    var hasNewline = chunk.indexOf('\n')!==-1;
    if(hasNewline){
        var arr = chunk.split('\n');
        var first = arr.shift();
        var last = arr.pop();
        data+=first;
        put(data);          //write a complete line
        arr.forEach(function(line){
            put(line);      //write a complete line
        });
        data=last;
    }else{
        data+=chunk;
    }
    if(n>100){
        end();
    }
}
function onErr(e){
    console.log(e);
}

r.addListener( "data", onData);
r.addListener( "end", end);
r.addListener('error',onErr);
w.addListener('error',onErr);

我看到了两个问题

第一个是
end
函数调用ReadStream上的
destroy
，但在一般情况下，这是由
end
事件触发的，这意味着流已经关闭，它将自动调用
destroy
。这意味着
r.destroy
将被调用两次，从而触发一个错误。这就是您看到的打印错误的原因

第二个问题是您正在WriteStream上调用
destroy
。我建议你去阅读相关文件：

具体地说，
任何排队的写入数据都不会被发送
，这就是您丢失一些输出的原因

基本上，如果您希望ReadStream提前关闭，您应该只调用ReadStream上的
destroy
，就像您的
n>100
案例一样。然后您希望使用WriteStream的
end
，以便流有时间写入所有缓冲数据

这是一个简化的版本，我认为应该是一样的。我也不会麻烦绑定
error
，因为不管怎样，错误都会自动打印到控制台

var fs = require('fs');
var data ='';
var n=0;                    //For line control

var r = fs.createReadStream('./input.txt',{
    encoding: 'ascii',
    start:0,
    // end: 100000,
});

var w = fs.createWriteStream('./output.txt',{
    encoding:'ascii'
});

r.addListener( "data", function(chunk){
    data += chunk;
    var lines = data.split('\n');
    data = lines.pop();

    lines.forEach(function(line){
      if (!r.readable) return; // If already destroyed
      if (n >= 100) {
          // Stop any more 'data' events and close the file.
          // This will also trigger 'close' below and close the writestream.
          r.destroy();
          return;
      }

      n++;
      w.write(line + '\n');
    });
});
r.addListener( "end", function(){
    // When we hit the end of the file, close the write stream,
    // and write any remaining line content
    w.write(data);
});
r.addListener("close", function(){
  w.end();
});

啊，行得通。谢谢，这看起来像是一个典型的新手错误。