You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@couchdb.apache.org by Nathan Vander Wilt <na...@calftrail.com> on 2010/09/16 06:59:13 UTC
Binary form fields versus update handlers
Hello all,
Tonight I decided I'd see how quickly I could implement a custom image upload site for The iPhone App Formerly Known As Tweetie (http://developer.atebits.com/tweetie-iphone/custom-image/) using a CouchDB update handler.
Unfortunately for me, Twitter.app sends the image as binary data:
'--com.aTeBiTs.TwEeTiE.oeunht43353r4cg345nt\r\nContent-Disposition: form-data; name="message"\r\n\r\nTest \r\n--com.aTeBiTs.TwEeTiE.oeunht43353r4cg345nt\r\nContent-Disposition: form-data; name="media"; filename="image.jpg"\r\nContent-Type: image/jpeg\r\n\r\n\xff\xd8' + more_binary_data
...and this causes mochiweb unhappiness:
{mochiweb_socket_server,235,{child_error,{ucs,{bad_utf8_character_code}}}}
I'm assuming MochiWeb is able to handle binary POST data, and this exception is just due to CouchDB wanting to convert the request to something JavaScript can handle. But beyond that (and that's a pretty lousy starting point, eh?) I'm wondering if and how we can make things like my little project possible.
In this case, my update handler needs to be able to put some of the form parts/fields (which are UTF-8able) into a document and another of the form fields (the one with the binary) as an attachment to a document. Would it be reasonable for Couch to pull out binary form parts (detected in some suitable fashion) in some fashion? Perhaps a separate req.attachments structure (with base64 encoding already done, naturally)? I'm no RFC2388 expert (at least not yet :-) but I wanted to get this out for discussion as a first step.
thanks,
-natevw
Re: Binary form fields versus update handlers
Posted by Chris Anderson <jc...@apache.org>.
On Wed, Sep 15, 2010 at 9:59 PM, Nathan Vander Wilt
<na...@calftrail.com> wrote:
> Hello all,
>
> Tonight I decided I'd see how quickly I could implement a custom image upload site for The iPhone App Formerly Known As Tweetie (http://developer.atebits.com/tweetie-iphone/custom-image/) using a CouchDB update handler.
>
> Unfortunately for me, Twitter.app sends the image as binary data:
>
> '--com.aTeBiTs.TwEeTiE.oeunht43353r4cg345nt\r\nContent-Disposition: form-data; name="message"\r\n\r\nTest \r\n--com.aTeBiTs.TwEeTiE.oeunht43353r4cg345nt\r\nContent-Disposition: form-data; name="media"; filename="image.jpg"\r\nContent-Type: image/jpeg\r\n\r\n\xff\xd8' + more_binary_data
>
> ...and this causes mochiweb unhappiness:
> {mochiweb_socket_server,235,{child_error,{ucs,{bad_utf8_character_code}}}}
>
> I'm assuming MochiWeb is able to handle binary POST data, and this exception is just due to CouchDB wanting to convert the request to something JavaScript can handle. But beyond that (and that's a pretty lousy starting point, eh?) I'm wondering if and how we can make things like my little project possible.
>
> In this case, my update handler needs to be able to put some of the form parts/fields (which are UTF-8able) into a document and another of the form fields (the one with the binary) as an attachment to a document. Would it be reasonable for Couch to pull out binary form parts (detected in some suitable fashion) in some fashion? Perhaps a separate req.attachments structure (with base64 encoding already done, naturally)? I'm no RFC2388 expert (at least not yet :-) but I wanted to get this out for discussion as a first step.
I like your proposed solution (the magic base64). It will probably be
easy enough to patch based on otherstuff happening in CouchDB (look at
the attachment POST api). I dont' have time these days to hack
something like this, but you have my full encouragement.
Chris
>
> thanks,
> -natevw
--
Chris Anderson
http://jchrisa.net
http://couch.io