PhantomJS 2.0 webserver module response.setEncoding("binary") not working

[leoweigand]

In phantomjs 2.0, calling setEncoding("binary") on the response object has no effect. What was working fine with version 1.9.7_1 now seems to be broken.
Code to replicate (where test.pdf may be any other binary file in the directory of the script):

var server  = require('webserver').create(),
        fs  = require('fs');

server.listen(8000, function(request, response) {
    response.statusCode = 200;
    response.setHeader('Content-type', 'application/pdf')
    response.setEncoding('binary');

    response.write( fs.read('test.pdf', 'b') );
    response.close();   
});

I am running on Mac OSX 10.10.

[ghost]

I can also confirm that this is true. Multiple hours were spent wondering why examples wouldn't work so thank you @lweigand :D. Works fine on 1.9.8 on Windows.

[Verikon]

Confirmed this is broken. (ubuntu, compiled Phantom2 from source).

[GoldHornKing]

Confirmed in windows. It worked with version 1.9.8.

The pdf file downloaded from response cannot be opened, Adobe Reader says it is damaged.

I also found version 2.0 generates large PDF than 1.9.8 for same page.

[smalltalking]

It looks like the content is encoded to UTF-8 - something that doesn't happen with 1.9.7 - then it gets truncated to the length before the encoding, so the end of the file is lost.

[griffinmyers]

+1 big bummer

[jigfox]

confirmed on mac, too

downgrading to phantomjs 1.9.8 helped solving this issue

[marcosbrigante]

Same here. Any workarounds to avoid downgrading?

[griffinmyers]

@marcosbrigante we pushed the encoding down into the client, i.e. in the response handler from phantom did something like this:

 var binaryString = new Buffer(response).toString('utf-8');
 return new Buffer(binaryString, 'binary');

[marcosbrigante]

@griffinmyers thanks I will give it a try. My current workaround is to use base64 instead of binary to transfer the pdf. But it requires decoding on the client...

response.setHeader('Content-type', 'application/base64');
response.write(btoa(fs.read('test.pdf', 'b')));

[GoldHornKing]

For anyone still stuck here, please try this work around. Thanks for @marcosbrigante and all other posts.

PhantomJS script:

var pdfStream = fs.open(fileLocation, 'rb'); 
var pdfContent = pdfStream.read();              
var base64Encoded = btoa(pdfContent);
response.statusCode = 200;
response.headers = {
                'Cache': 'no-cache',
                'Content-Type': 'application/base64',
                'Connection': 'Keep-Alive',
                'Content-Length': base64Encoded.length.toString()
                };
response.write(base64Encoded);
response.close();

Server side [C#]:

HttpWebResponse response =     (HttpWebResponse)httpWReq.GetResponse();
using (Stream stream = response.GetResponseStream())
{
    using (StreamReader readStream = new StreamReader(stream))
    {
        string base64String = readStream.ReadToEnd();
        byte[] decodedArray = Convert.FromBase64String(base64String);

        // The following three lines are testing if content is a valid pdf
        using (FileStream stream1 = System.IO.File.Create("c:\\test.pdf"))
        {
            stream1.Write(decodedArray, 0, decodedArray.Length);
        }
        return decodedArray;
}

//decodedArray could be then added in a  HttpRequest

[jigfox]

Is there a timeline, when this will be released?

[rcarmo]

I'm still having trouble with this, even with 2.1. Is that ".toLatin()" conversion right?

[vitallium]

@rcarmo yes. It works. Just tested it.

[rcarmo]

for me, this:

res.setEncoding('binary');
res.write(fs.open(path,'rb').read());
res.close();

Is still failing under 2.1. Trying to figure out why the webserver doesn't like it...

[kkeybbs]

In an old version(2.0.0-windows), setEncoding('binary') doesn't work.but rsp.setEncoding('latin1') works.

I update to the 2.1.1-windows, setEncoding('binary') works too.

So you can use setEncoding('latin1') if you don't want to update at once.