12

According to the CouchDB Wiki on PUT operations.

To update an existing document, you also issue a PUT request. In this case, the JSON body must contain a _rev property, which lets CouchDB know which revision the edits are based on. If the revision of the document currently stored in the database doesn't match, then a 409 conflict error is returned.

My goal is to perform a bulk_docs update:

curl -X POST [domain]/[couch db name]/_bulk_docs -H "Content-type: application/json" -d @[some document].json

My workflow is like this:

  1. My data is in an Google Docs spreadsheet.
  2. I convert the spreadsheet data to JSON by copying and pasting into Mr. Data Converter
  3. I use cURL (as show above) to add/update documents

The problem is that the first time I add new documents, everything works perfectly, however the next time I post the same documents, I get following error for each document:

...{"id":"28"," error":"conflict","reason":"Document update conflict."}...

Is there any way to update an existing document without including a _rev property?

Octavian Helm
  • 39,405
  • 19
  • 98
  • 102
edt
  • 22,010
  • 30
  • 83
  • 118

1 Answers1

21

By design, you cannot update a CouchDB document blindly, you can only attempt to update a specific revision of a document.

For a single document, you can use a CouchDB update handler to hide this from the client as an update handler will be passed the existing document (if it exists) including its revision.

For a collection of documents, when using _bulk_docs, you can add "new_edits": false which will forcibly insert conflicts instead of rejection (though you'll still need to pass a _rev, it just doesn't have to be the current one).

All that said, it would be better to follow the rules. Grab the current revision of the document you would like to update, attempt to update it, if you get a 409, get the new version, merge as appropriate, and update again.

Octavian Helm
  • 39,405
  • 19
  • 98
  • 102
Robert Newson
  • 4,631
  • 20
  • 18
  • The update handler seems to be the way to go for updating data. It will be faster because you don't have to download every document, merge it client side, and upload it again. Instead the update handler can do the merging for you with 1/3rd of the network activity! – Dobes Vandermeer Nov 22 '11 at 03:21
  • @DobesVandermeer that is not universal. Since update handler does not support batch ops, that's one request per update/persist. If you need to do more than 2-4 of them simultaneously, do batch operations because it usually takes 2-4 requests depending on circumstances to for example update 10 docs. Of course if you use `"new_edits": false` you can batch persist/replace in 1 request, but be aware, that `new_edits` is [intended to be used by the replicator](http://goo.gl/EFFofn). A drawback is that it doesn't preserve revision history, I'm wondering if that might interfere with replication... – Dominykas Mostauskis Feb 20 '14 at 11:13
  • @DominykasMostauskis ah yes, good point. The OP might want to do a batch download, calculate the new values, and batch upload which is fewer transactions overall. As long as it all fits in memory. – Dobes Vandermeer Feb 21 '14 at 18:02
  • If I'm not mistaken, using new_edits:false without the current _rev will result in a conflict, and which doc is chosen during automatic conflict resolution is up to Couch. So until you go in and fix the conflicts yourself, you'll see some of the bulk inserted docs, but not all of them, because they are being hidden by a conflicting rev. – Rich Remer Aug 21 '14 at 16:25
  • 1
    Upvote because you mentioned `you'll still need to pass a _rev`, that helped me a lot! – Per Lundberg Mar 30 '17 at 10:29