Stucks sometimes on piping stage in the middle of stream about readabilitysax HOT 12 CLOSED

I know you have another request wrapper thing which probably doesn't triggers this bug, however I need some more features than that and superagent seems to be the best one. I am piping from superagent to lots of different modules and never seen a problem like this ... though probably there is something with WritableStream ... you might want to look at event-stream module, which works pretty much always f.e. the .through method ... probably they have fixed something like this.

from readabilitysax.

I haven't tried to understand whats the reason, I just took event-stream and replaced WritableStream by this, now it works perfectly

var Readability = readabilitySax.Readability
var Parser = htmlParser.Parser
var CollectingHandler = htmlParser.CollectingHandler

var readability = new Readability({pageURL: url, type: 'html'})
var handler = new CollectingHandler(readability)
var parser = new Parser(handler, {lowerCaseTags: true})

return es.through(function(data) {
    parser.write(data)
    this.emit('data', data)
}, function() {
    for(
        var skipLevel = 1;
        readability._getCandidateNode().info.textLength < 250 && skipLevel < 4;
        skipLevel++
    ){
        readability.setSkipLevel(skipLevel)
        handler.restart()
    }

    var article = readability.getArticle()
    article.html = entities.decodeHTML5(article.html.replace(/\s+/g, ' '))
    article.title = entities.decodeHTML5(article.title)
    article.url = url
    callback(article)
    this.emit('end')
})

from readabilitysax.

readability._getCandidateNode() this is also not nice, you are accessing private or protected method from the outside.

from readabilitysax.

The stream interface apparently has a bug. A PR replacing it (eg. using through2) would be great; your code should be trivial to port, although entities could be replaced with the decodeEntities option of htmlparser2.

from readabilitysax.

What do you think exactly is a bug?

from readabilitysax.

The end method currently overwrites the original end method of the stream interface, which could be the root of this issue.

from readabilitysax.

How is the end related to the fact that streaming stops at some point in the middle?

from readabilitysax.

I found out that this code works without issues:

var request = require('superagent')
var readabilitySax = require('readabilitySAX')

var url = 'http://techcrunch.com/2010/11/18/mark-zuckerberg/'

var options = {pageURL: url, type: 'html'}

request
    .get(url)
    .on('error', console.log)
    .pipe(readabilitySax.createWritableStream(options, console.log))

from readabilitysax.

weird ...

from readabilitysax.

^^

The stream implementation is buggy nevertheless & should probably be updated. I'm closing this issue though.

from readabilitysax.

I have switched to full buffering ... streams the way I have used them were a way slower and fragile.

from readabilitysax.

also encoding detection can be done safely only if the full text is given.

from readabilitysax.