读取 n 行大文本文件

Question

我拥有的最小文件有 > 850k 行，每行的长度未知。目标是在浏览器中从此文件中读取n 行。完全阅读它是不可能的。

这是 HTML <input type="file" name="file" id="file"> 和我拥有的 JS：

var n = 10;
var reader = new FileReader();
reader.onload = function(progressEvent) {
  // Entire file
  console.log(this.result);

  // By lines
  var lines = this.result.split('\n');
  for (var line = 0; line < n; line++) {
    console.log(lines[line]);
  }
};

很明显，这里的问题是它首先尝试真实的整个文件，然后用换行符分割它。所以不管n，它都会尝试读取整个文件，最终当文件很大时什么都不读取。

我该怎么做？

注意：我愿意删除整个函数并从头开始，因为我将能够console.log() 我们阅读的每一行。

---

*"每一行的长度未知" -> 表示文件是这样的：

(0, (1, 2))
(1, (4, 5, 6))
(2, (7))
(3, (8))

---

编辑：

要走的路类似于filereader api on big files，但我不知道如何修改它以读取文件的n 行...

也可以使用Uint8Array to string in Javascript，从那里开始：

var view = new Uint8Array(fr.result);
var string = new TextDecoder("utf-8").decode(view);
console.log("Chunk " + string);

但这可能不会将最后一行作为一个整体读取，那么您以后将如何确定这些行呢？例如这里是它打印的内容：

((7202), (u'11330875493', u'2554375661'))
((1667), (u'9079074735', u'6883914476',

Answer 1

我需要在浏览器中读取 250MB 的 utf-8 编码文件。我的解决方案是编写类似 TextReader 类的 C#，它给了我类似异步流的行为。

---

TextReader 类：

class TextReader {
    CHUNK_SIZE = 8192000; // I FOUND THIS TO BE BEST FOR MY NEEDS, CAN BE ADJUSTED
    position = 0;
    length = 0;

    byteBuffer = new Uint8Array(0);

    lines = [];
    lineCount = 0;
    lineIndexTracker = 0;

    fileReader = new FileReader();
    textDecoder = new TextDecoder(`utf-8`);

    get allCachedLinesAreDispatched() {
        return !(this.lineIndexTracker < this.lineCount);
    }

    get blobIsReadInFull() {
        return !(this.position < this.length);
    }

    get bufferIsEmpty() {
        return this.byteBuffer.length === 0;
    }

    get endOfStream() {
        return this.blobIsReadInFull && this.allCachedLinesAreDispatched && this.bufferIsEmpty;
    }

    constructor(blob) {
        this.blob = blob;
        this.length = blob.size;
    }

    blob2arrayBuffer(blob) {
        return new Promise((resolve, reject) => {
            this.fileReader.onerror = reject;
            this.fileReader.onload = () => {
                resolve(this.fileReader.result);
            };

            this.fileReader.readAsArrayBuffer(blob);
        });
    }

    read(offset, count) {
        return new Promise(async (resolve, reject) => {
            if (!Number.isInteger(offset) || !Number.isInteger(count) || count < 1 || offset < 0 || offset > this.length - 1) {
                resolve(new ArrayBuffer(0));
                return
            }

            let endIndex = offset + count;

            if (endIndex > this.length) endIndex = this.length;

            let blobSlice = this.blob.slice(offset, endIndex);

            resolve(await this.blob2arrayBuffer(blobSlice));
        });
    }

    readLine() {
        return new Promise(async (resolve, reject) => {

            if (!this.allCachedLinesAreDispatched) {
                resolve(this.lines[this.lineIndexTracker++] + `\n`);
                return;
            }

            while (!this.blobIsReadInFull) {
                let arrayBuffer = await this.read(this.position, this.CHUNK_SIZE);
                this.position += arrayBuffer.byteLength;

                let tempByteBuffer = new Uint8Array(this.byteBuffer.length + arrayBuffer.byteLength);
                tempByteBuffer.set(this.byteBuffer);
                tempByteBuffer.set(new Uint8Array(arrayBuffer), this.byteBuffer.length);

                this.byteBuffer = tempByteBuffer;

                let lastIndexOfLineFeedCharacter = this.byteBuffer.lastIndexOf(10); // LINE FEED CHARACTER (\n) IS ONE BYTE LONG IN UTF-8 AND IS 10 IN ITS DECIMAL FORM

                if (lastIndexOfLineFeedCharacter > -1) {
                    let lines = this.textDecoder.decode(this.byteBuffer).split(`\n`);
                    this.byteBuffer = this.byteBuffer.slice(lastIndexOfLineFeedCharacter + 1);

                    let firstLine = lines[0];

                    this.lines = lines.slice(1, lines.length - 1);
                    this.lineCount = this.lines.length;
                    this.lineIndexTracker = 0;

                    resolve(firstLine + `\n`);
                    return;
                }
            }

            if (!this.bufferIsEmpty) {
                let line = this.textDecoder.decode(this.byteBuffer);
                this.byteBuffer = new Uint8Array(0);
                resolve(line);
                return;
            }

            resolve(null);
        });
    }
}

---

用法：

document.getElementById("read").onclick = async () => {
    let file = document.getElementById("fileInput").files[0];
    let textReader = new TextReader(file);

    while(true) {
        let line = await textReader.readLine();
        if(line === null) break;
        // PROCESS LINE
    }

    // OR

    while (!textReader.endOfStream) {
        let line = await textReader.readLine();
        // PROCESS LINE
    }
};

---

性能：

我能够在 1.5 秒内读取包含 1,398,258 行的单个 250MB utf-8 编码文本文件，JS 堆大小不超过 20MB。相比之下，如果我一次性读取同一个文件，然后将结果字符串按 \n 拆分，仍然需要 ~1.5 秒，但是 JS Heap 会达到 230MB。

Answer 2

流是特色！
whatwg 团队正在研究关于可写 + 可读流的最后一个变化，并且很快就准备好了。但在那之前，您可以使用web-stream-polyfill。 他们正在研究一种从 blob 以及 [1] 获取 ReadableStream 的方法。但我还创建了一种以流方式获取 blob 的方法：Screw-FileReader

昨天我还创建了一个node-byline 的简单port 来处理网络流

所以这可以很简单：

// Simulate a file
var csv =
`apple,1,$1.00
banana,4,$0.20
orange,3,$0.79`

var file = new Blob([csv])

var n = 0
var controller
var decoder = new TextDecoder
var stdout = new WritableStream({
  start(c) {
      controller = c
    },
    write(chunk, a) {
      // Calling controller.error will also put the byLine in an errored state
      // Causing the file stream to stop reading more data also
      if (n == 1) controller.error("don't need more lines")
      chunk = decoder.decode(chunk)
      console.log(`chunk[${n++}]: ${chunk}`)
    }
})

file
  .stream()
  .pipeThrough(byLine())
  // .pipeThrough(new TextDecoder) something like this will work eventually
  .pipeTo(stdout)

<script src="https://cdn.rawgit.com/creatorrr/web-streams-polyfill/master/dist/polyfill.min.js"></script>
<script src="https://cdn.rawgit.com/jimmywarting/Screw-FileReader/master/index.js"></script>

<!-- after a year or so you only need byLine -->
<script src="https://cdn.rawgit.com/jimmywarting/web-byline/master/index.js"></script>

Answer 3

逻辑与我在对filereader api on big files 的回答中所写的非常相似，只是您需要跟踪到目前为止已处理的行数（以及到目前为止读取的最后一行，因为它可能还没有结束）。下一个示例适用于与 UTF-8 兼容的任何编码；如果您需要其他编码，请查看 TextDecoder 构造函数的选项。

如果您确定输入是 ASCII（或任何其他单字节编码），那么您也可以跳过使用 TextDecoder 并使用 FileReader's readAsText method 直接将输入读取为文本。

// This is just an example of the function below.
document.getElementById('start').onclick = function() {
    var file = document.getElementById('infile').files[0];
    if (!file) {
        console.log('No file selected.');
        return;
    }
    var maxlines = parseInt(document.getElementById('maxlines').value, 10);
    var lineno = 1;
    // readSomeLines is defined below.
    readSomeLines(file, maxlines, function(line) {
        console.log("Line: " + (lineno++) + line);
    }, function onComplete() {
        console.log('Read all lines');
    });
};

/**
 * Read up to and including |maxlines| lines from |file|.
 *
 * @param {Blob} file - The file to be read.
 * @param {integer} maxlines - The maximum number of lines to read.
 * @param {function(string)} forEachLine - Called for each line.
 * @param {function(error)} onComplete - Called when the end of the file
 *     is reached or when |maxlines| lines have been read.
 */
function readSomeLines(file, maxlines, forEachLine, onComplete) {
    var CHUNK_SIZE = 50000; // 50kb, arbitrarily chosen.
    var decoder = new TextDecoder();
    var offset = 0;
    var linecount = 0;
    var linenumber = 0;
    var results = '';
    var fr = new FileReader();
    fr.onload = function() {
        // Use stream:true in case we cut the file
        // in the middle of a multi-byte character
        results += decoder.decode(fr.result, {stream: true});
        var lines = results.split('\n');
        results = lines.pop(); // In case the line did not end yet.
        linecount += lines.length;
    
        if (linecount > maxlines) {
            // Read too many lines? Truncate the results.
            lines.length -= linecount - maxlines;
            linecount = maxlines;
        }
    
        for (var i = 0; i < lines.length; ++i) {
            forEachLine(lines[i] + '\n');
        }
        offset += CHUNK_SIZE;
        seek();
    };
    fr.onerror = function() {
        onComplete(fr.error);
    };
    seek();
    
    function seek() {
        if (linecount === maxlines) {
            // We found enough lines.
            onComplete(); // Done.
            return;
        }
        if (offset !== 0 && offset >= file.size) {
            // We did not find all lines, but there are no more lines.
            forEachLine(results); // This is from lines.pop(), before.
            onComplete(); // Done
            return;
        }
        var slice = file.slice(offset, offset + CHUNK_SIZE);
        fr.readAsArrayBuffer(slice);
    }
}

Read <input type="number" id="maxlines"> lines from
<input type="file" id="infile">.
<input type="button" id="start" value="Print lines to console">