Node.js：复制 csv 时文件大小的差异

Question

I have the below code where I am reading from a CSV and writing to another CSV. I will be transforming some data before writing to another file, but as a test, I ran the code and see that there are slight differences between source and destination files without event changing anything about the files.

  for(const m of metadata) {
      tempm = m;
      fname = path;
      const pipelineAsync = promisify(pipeline);
      if(m.path) {
        await pipelineAsync(
          fs.createReadStream(m.path),
          csv.parse({delimiter: '\t', columns: true}),
          csv.transform((input) => {
            return Object.assign({}, input);
          }),
          csv.stringify({header: true, delimiter: '\t'}),
          fs.createWriteStream(fname, {encoding: 'utf16le'})
        )
        let nstats = fs.statSync(fname);
        tempm['transformedPath'] = fname;
        tempm['transformed'] = true;
        tempm['t_size_bytes'] = nstats.size;
      }
  }

I see that for example,

file a: the source file size is `895631` while after copying destination file size is `898545`
file b: the source file size is `51388` while after copying destination file size is `52161`
file c: the source file size is `13666` while after copying destination file size is `13587`

But when i do not use tranform, the sizes match, for example this code produces excatly same file sizes on both source and dest

  for(const m of metadata) {
      tempm = m;
      fname = path;
      const pipelineAsync = promisify(pipeline);
      if(m.path) {
        await pipelineAsync(
          fs.createReadStream(m.path),
          /*csv.parse({delimiter: '\t', columns: true}),
          csv.transform((input) => {
            return Object.assign({}, input);
          }),
          csv.stringify({header: true, delimiter: '\t'}),*/
          fs.createWriteStream(fname, {encoding: 'utf16le'})
        )
        let nstats = fs.statSync(fname);
        tempm['transformedPath'] = fname;
        tempm['transformed'] = true;
        tempm['t_size_bytes'] = nstats.size;
      }
  }

Can any one please help in identifying what options i need to pass to csv transformation, so that the copy happens correctly.

I am doing this test to ensure, i am not losing out any data in large files.

Thanks.

Update 1: I have also checked that the encoding on both the files is same.

Update 2: I notice that the the source file has CRLF and destination file has LF . Is there a way i can keep the same using node.js or is it something OS dependent.

Update 3: Looks like the issue is EOL , I see the source file has CRLF while the destination file / transformed file has LF . I need to now find a way to specify this my above code so that the EOL is consistent

Answer 1

You need to setup you EOL config:

const { pipeline } = require('stream')
const { promisify } = require('util')
const fs = require('fs')
const csv = require('csv')
const os = require('os')


;(async function () {
  const pipelineAsync = promisify(pipeline)
  await pipelineAsync(
    fs.createReadStream('out'),
    csv.parse({ delimiter: ',', columns: true }),
    csv.transform((input) => {
      return Object.assign({}, input)
    }),
    // Here the trick:
    csv.stringify({ eol: true, record_delimiter: os.EOL, header: true, delimiter: '\t' }),
    fs.createWriteStream('out2', { encoding: 'utf16le' })
  )
})()

You can use \r\n as well or whatever you need $-new-line\n

This setup can be spotted reading the source code .

Answer 2

The two main source of this kind of difference are:

1. EOL style (unix or ms-dos)
2. file encoding

Using the simple unix file command line utility you can check both the encoding and EOL style for source files. Make sure to use same options for dest files and any difference should disappear.

Hope this helps.