[att] attachments are returned as strings about nano

[dscape]

gist me some code here please

[dscape]

If this doesnt break api and other clients, go ahead and commit

[pgte]

The current behaviour is just wrong.

In https://github.com/dscape/nano/blob/master/nano.js#L512 you pass the encoding as "binary". This is just wrong because, according to the Node docs:

"binary": A way of encoding raw binary data into strings by using only the first 8 bits of each character. This encoding method is deprecated and should be avoided in favor of Buffer objects where possible. This encoding will be removed in future versions of Node.

You should force {encoding: null} in request arguments, so a buffer is returned.

[pgte]

Will you accept a pull request with this change? :)

[dscape]

If it doesn't break tests yes. Also it shouldn't break current behavior.

@perezd @mikeal opinions?

[dscape]

Also does it work in expensecat with the changes you proposed? Does the pipe example in samples work?

[pgte]

It works for expensecat by doing the change.
Current behavior seams plain wrong, this is not usable IMO.
I got a file as a string, saved it and was totally borked, which is understandable given the given encoding.

[dscape]

@pgte have you tried it in 0.4.12. I think it will work there, but it's probably broken in 0.6.

Fix needs to pass in both test sets.

Also make sure this is being tested. It should, but just in case! I'm wondering why current tests pass???

https://github.com/dscape/nano/blob/master/tests/att/pipe.js#L8

If what you are saying is true, it shoudn't?

[dscape]

Maybe try run tests without nock. How can I make it not use nock?

NOCK_OUT=true node mytests.js

?

[pgte]

@dscape I'm using 0.4.11.

[dscape]

Shouldn't that test case be broken then?

We should meet and discuss this in detail.

Btw you are a repo collab, you can pretty much commit upstream as you like
(even though I'm sure companies using nano appreciate being warned about
these issues and following the discussion here)

[dscape]

@pgte how can i do this?

NOCK_OUT=true node mytests.js

I need to test this with iris couch

[dscape]

Pedro ping me on this monday or so so we can discuss / patch

[pgte]

Sorry, too busy, I'll ping you monday.

No dia 15/01/2012, às 01:30, Nuno Job
[email protected]
escreveu:

Pedro ping me on this monday or so so we can discuss / patch

---

Reply to this email directly or view it on GitHub:
#49 (comment)

[pgte]

For some reason nano attachments are working when they're piped, but not when they're being transfered in bulk.

For testing this I created a local database named "tests" and created this object and attached "js.png"

{
   "_id": "60a83e3b23fe83a873105c0ef1000ad7",
   "_rev": "2-2c4296b27533399ec1690eaf154f4950",
   "a": 1,
   "_attachments": {
       "js.png": {
           "content_type": "image/png",
           "revpos": 2,
           "length": 10268,
           "stub": true
       }
   }
}

Then I ran this code:

var fs = require('fs')
  , db = require('nano')('http://127.0.0.1:5984').use('tests');

function error(err) {
  throw err;
}

// ------- piping
(function() {
  var rs = db.attachment.get('60a83e3b23fe83a873105c0ef1000ad7', 'js.png');
  rs.on('error', error);
  var ws = fs.createWriteStream(__dirname + '/js.piped.png');
  ws.on('error', error);

  rs.pipe(ws);
}());

// ------- getting
(function() {
  db.attachment.get('60a83e3b23fe83a873105c0ef1000ad7', 'js.png', function(err, image) {
    if (err) { throw err; }
    fs.writeFile(__dirname + '/js.got.png', image, function(err) {
      if (err) { throw err; }
    });
  });
}());

It outputs 2 files:

• js.piped.png
• js.got.png

If you run this you will see the output files:

pedroteixeira:temp pedroteixeira$ ls -la *.png
-rw-r--r--  1 pedroteixeira  staff  14678 Jan 16 09:50 js.got.png
-rw-r--r--  1 pedroteixeira  staff  10268 Jan 16 09:50 js.piped.png

You can see that the sizes are different. Also, the js.got.png is corrupted:

pedroteixeira:temp pedroteixeira$ file js.piped.png 
js.piped.png: PNG image data, 1052 x 1052, 8-bit colormap, non-interlaced
pedroteixeira:temp pedroteixeira$ file js.got.png 
js.got.png: data

[dscape]

Go ahead and submit the pull request. Also add that one that fails to the test case.

I can do this myself but didn't cause I suppose you already have it :)

[dscape]

Also how can we nock this ?

[dscape]

Was this fixed?

[thiagoarrais]

I've tried to fix this in my attachment-buffer branch. I was just messing around to figure out if this was doable, so I didn't take a look at the tests. But the fix I applied works with @pgte's sample.

The relevant commit is 9ba2f2

[dscape]

@pgte can you please review? I never had this issue

[pgte]

@dscape Yes, that's exactly what my fix was that I had in a branch somewhere.
Looks good to me.

[thiagoarrais]

So I guess the problem boils down to backwards compatibility testing. I'll get to it when I have some time (which, admittedly, may turn out to be some weeks), but shouldn't we consider working with buffers even if it breaks backwards compatibility? Strings and encoding for binary files seem to be getting deprecated in node, and buffers can be encoded/decoded to/from strings if needed anyway.

[thiagoarrais]

Just a little addendum as to the evidence of Buffers being preferred instead of binary encoded strings (which is the solution we're using in nano right now): take a look at the description for the binary encoding at http://nodejs.org/api/buffer.html#buffer_buffer

[dscape]

Test Run is available here:

• http://travis-ci.org/#!/dscape/nano/builds/813212

If you guys can add the tests that were failing I would really appreciate it.

Nuno

[thiagoarrais]

I've taken the liberty to add a test for attachments as buffers to the main master. Feel free to revert if you see the need.

[dscape]

Great stuff, thanks Thiago :)