Parse multi-part formdata in the browser

multipart.js

var Multipart = {
  parse: (function() {
    function Parser(arraybuf, boundary) {
      this.array = arraybuf;
      this.token = null;
      this.current = null;
      this.i = 0;
      this.boundary = boundary;
    }

    Parser.prototype.skipPastNextBoundary = function() {
      var boundaryIndex = 0;
      var isBoundary = false;

      while (!isBoundary) {
        if (this.next() === null) {
          return false;
        }

        if (this.current === this.boundary[boundaryIndex]) {
          boundaryIndex++;
          if (boundaryIndex === this.boundary.length) {
            isBoundary = true;
          }
        } else {
          boundaryIndex = 0;
        }
      }

      return true;
    }

    Parser.prototype.parseHeader = function() {
      var header = '';
      var _this = this;
      var skipUntilNextLine = function() {
        header += _this.next();
        while (_this.current !== '\n' && _this.current !== null) {
          header += _this.next();
        }
        if (_this.current === null) {
          return null;
        }
      };

      var hasSkippedHeader = false;
      while (!hasSkippedHeader) {
        skipUntilNextLine();
        header += this.next();
        if (this.current === '\r') {
          header += this.next(); // skip
        }

        if (this.current === '\n') {
          hasSkippedHeader = true;
        } else if (this.current === null) {
          return null;
        }
      }

      return header;
    }

    Parser.prototype.next = function() {
      if (this.i >= this.array.byteLength) {
        this.current = null;
        return null;
      }

      this.current = String.fromCharCode(this.array[this.i]);
      this.i++;
      return this.current;
    }

    function buf2String(buf) {
      var string = '';
      buf.forEach(function (byte) {
        string += String.fromCharCode(byte);
      });
      return string;
    }

    function processSections(arraybuf, sections) {
      for (var i = 0; i !== sections.length; ++i) {
        var section = sections[i];
        if (section.header['content-type'] === 'text/plain') {
          section.text = buf2String(arraybuf.slice(section.bodyStart, section.end));
        } else {
          var imgData = arraybuf.slice(section.bodyStart, section.end);
          section.file = new Blob([imgData], {
            type: section.header['content-type']
          });
          var fileNameMatching = (/\bfilename\=\"([^\"]*)\"/g).exec(section.header['content-disposition']) || [];
          section.fileName = fileNameMatching[1] || '';
        }
        var matching = (/\bname\=\"([^\"]*)\"/g).exec(section.header['content-disposition']) || [];
        section.name = matching[1] || '';

        delete section.headerStart;
        delete section.bodyStart;
        delete section.end;
      }

      return sections;
    }

    function multiparts(arraybuf, boundary) {
      boundary = '--' + boundary;
      var parser = new Parser(arraybuf, boundary);

      var sections = [];
      while (parser.skipPastNextBoundary()) {
        var header = parser.parseHeader();

        if (header !== null) {
          var headerLength = header.length;
          var headerParts = header.trim().split('\n');

          var headerObj = {};
          for (var i = 0; i !== headerParts.length; ++i) {
            var parts = headerParts[i].split(':');
            headerObj[parts[0].trim().toLowerCase()] = (parts[1] || '').trim();
          }

          sections.push({
            'bodyStart': parser.i,
            'header': headerObj,
            'headerStart': parser.i - headerLength
          });
        }
      }

      // add dummy section for end
      sections.push({
        'headerStart': arraybuf.byteLength - 2 // 2 hyphens at end
      });
      for (var i = 0; i !== sections.length - 1; ++i) {
        sections[i].end = sections[i+1].headerStart - boundary.length;

        if (String.fromCharCode(arraybuf[sections[i].end]) === '\r' || '\n') {
          sections[i].end -= 1;
        }
        if (String.fromCharCode(arraybuf[sections[i].end]) === '\r' || '\n') {
          sections[i].end -= 1;
        }
      }
      // remove dummy section
      sections.pop();

      sections = processSections(arraybuf, sections);

      return sections;
    }

    return multiparts;
  })()
};

Ok fine thank you.

I've re-written it in TypeScript as a ES-module, but I've not tested it yet. It works fine in my case in Cypress anyway.

class Parser {
    private array: Uint8Array;
    // private token: null;
    private current: string | null;
    public i: number;
    private boundary: string;

    public constructor(arraybuf: Uint8Array, boundary: string) {
        this.array = arraybuf;
        // this.token = null;
        this.current = null;
        this.i = 0;
        this.boundary = boundary;
    }

    public skipPastNextBoundary(): boolean {
        let boundaryIndex = 0;
        let isBoundary = false;

        while (!isBoundary) {
            if (this.next() === null) {
                return false;
            }

            if (this.current === this.boundary[boundaryIndex]) {
                boundaryIndex++;
                if (boundaryIndex === this.boundary.length) {
                    isBoundary = true;
                }
            } else {
                boundaryIndex = 0;
            }
        }

        return true;
    }

    public parseHeader() {
        let header = '';
        const skipUntilNextLine = () => {
            header += this.next();
            while (this.current !== '\n' && this.current !== null) {
                header += this.next();
            }
            if (this.current === null) {
                return null;
            }
        };

        let hasSkippedHeader = false;
        while (!hasSkippedHeader) {
            skipUntilNextLine();
            header += this.next();
            if (this.current === '\r') {
                header += this.next(); // skip
            }

            if (this.current === '\n') {
                hasSkippedHeader = true;
            } else if (this.current === null) {
                return null;
            }
        }

        return header;
    }

    public next() {
        if (this.i >= this.array.byteLength) {
            this.current = null;
            return null;
        }

        this.current = String.fromCharCode(this.array[this.i]);
        this.i++;
        return this.current;
    }
}

function buf2String(buf: Uint8Array): string {
    return Array.from(buf)
        .map((byte) => String.fromCharCode(byte))
        .join('');
}

interface Section {
    header?: Record<string, string>;
    text?: string;
    file?: Blob;
    fileName?: string;
    name?: string;
    bodyStart?: number;
    end?: number;
    headerStart?: number;
}

function processSections(arraybuf: Uint8Array, sections: Section[]): Section[] {
    for (let i = 0; i !== sections.length; ++i) {
        const section = sections[i];
        if (section.header!['content-type'] === 'text/plain') {
            section.text = buf2String(arraybuf.slice(section.bodyStart, section.end));
        } else {
            const imgData = arraybuf.slice(section.bodyStart, section.end);
            section.file = new Blob([imgData], {
                type: section.header!['content-type'],
            });
            const fileNameMatching = /\bfilename="([^"]*)"/g.exec(section.header!['content-disposition']) || [];
            section.fileName = fileNameMatching[1] || '';
        }
        const matching = /\bname="([^"]*)"/g.exec(section.header!['content-disposition']) || [];
        section.name = matching[1] || '';

        delete section.headerStart;
        delete section.bodyStart;
        delete section.end;
    }

    return sections;
}

function multiparts(arraybuf: Uint8Array, boundary: string) {
    boundary = '--' + boundary;
    const parser = new Parser(arraybuf, boundary);

    let sections: Section[] = [];
    while (parser.skipPastNextBoundary()) {
        const header = parser.parseHeader();

        if (header !== null) {
            const headerLength = header.length;
            const headerParts = header.trim().split('\n');

            const headerObj: Record<string, string> = {};
            for (let i = 0; i !== headerParts.length; ++i) {
                const parts = headerParts[i].split(':');
                headerObj[parts[0].trim().toLowerCase()] = (parts[1] || '').trim();
            }

            sections.push({
                bodyStart: parser.i,
                header: headerObj,
                headerStart: parser.i - headerLength,
            });
        }
    }

    // add dummy section for end
    sections.push({
        headerStart: arraybuf.byteLength - boundary.length - 2, // 2 hyphens at end
    });
    for (let i = 0; i !== sections.length - 1; ++i) {
        sections[i].end = sections[i + 1].headerStart! - boundary.length;

        if (['\r', '\n'].includes(String.fromCharCode(arraybuf[sections[i].end!]))) {
            sections[i].end! -= 1;
        }
    }
    // remove dummy section
    sections.pop();

    sections = processSections(arraybuf, sections);

    return sections;
}

export interface ParsedSection {
    blob: Blob;
    fileName?: string;
}

export type ParseResult = Record<string, ParsedSection>;

export function parse(arraybuf: Uint8Array, boundary: string): ParseResult {
    return multiparts(arraybuf, boundary).reduce<ParseResult>((acc, section) => {
        acc[section.name!] = {
            blob: section.file!,
            fileName: section.fileName,
        };

        return acc;
    }, {});
}

There is bug in this code.
The following part is buggy,

sections.push({
        'headerStart': arraybuf.byteLength - boundary.length - 2 // 2 hyphens at end
      });

You are removing boundary length two times for last section. You already remove boundary when define section.end. So, fix is following

sections.push({
        'headerStart': arraybuf.byteLength - 2 // 2 hyphens at end
      });

Modern browsers can parse multipart/form-data natively. Example:

const payload =
`------WebKitFormBoundaryU5rJUDxGnj15hIGW\r
Content-Disposition: form-data; name="field1"\r
\r
Hello
World,
This is me\r
------WebKitFormBoundaryU5rJUDxGnj15hIGW--`

const boundary = payload.slice(2, payload.indexOf('\r\n'))
new Response(payload, {
  headers: {
    'Content-Type': `multipart/form-data; boundary=${boundary}`
  }
})
  .formData()
  .then(formData => {
    console.log([...formData]) // [['field1', 'Hello\nWorld,\nThis is me']]
  })

The \r inside payload are necessary, because the line breaks must be \r\n, except the values themselves. If you have a properly formed multipart/form-data blob, you don't need to add \r.

If you want to parse an HTTP response, you can use the fetch response directly:

fetch('/formdata-response')
  .then(response => response.formData())
  .then(formData => console.log([...formData]))

Modern browsers can parse multipart/form-data natively. Example:

Thank you! We've come a long way since 2017

@Finesse entries() is not necessary, FormData spreads to an array of arrays (entries) [...fd].

@guest271314 You are right, I've amended my code snippet

Very helpful snippet. If you have access to fetch() it should be possible to use text() to get the raw multipart/form-data content with \r\n included

{
  var formdata = new FormData();
  var dirname = "web-directory";
  formdata.append(dirname, new Blob(["123"], {
    type: "text/plain"
  }), `${dirname}/file.txt`);
  formdata.append(dirname, new Blob(["src"], {
    type: "text/plain"
  }), `${dirname}/src/file.txt`);
  var body = await new Response(formdata).text();
  console.log(body);
  const boundary = body.slice(2, body.indexOf('\r\n'))
  console.log(boundary)
  new Response(body, {
      headers: {
        'Content-Type': `multipart/form-data; boundary=${boundary}`
      }
    })
    .formData()
    .then(formData => {
      console.log([...formData])
    })
}

@Finesse Another way to do this when payload is a TypedArray (or ArrayBuffer)

let ab = new Uint8Array(await response.clone().arrayBuffer());

let boundary = ab.subarray(2, ab.indexOf(13) + 1);

let archive = await new Response(ab, {
    headers: {
      "Content-Type": `multipart/form-data; boundary=${new TextDecoder().decode(boundary)}`,
    },
  })
  .formData()
  .then((data) => {
    console.log([...data]);
    return data;
  }).catch((e) => {
    console.warn(e);
  });