Yes, that's true. An update_seq is always valid JSON and should be
treated opaquely, though a change does flush out bugs in client
libraries and such.

B.

On 15 January 2013 12:32, Jens Alfke <[email protected]> wrote:
> I just got a bug report about TouchDB failing to sync with BigCouch. From the 
> symptoms, it appears that BigCouch has recently changed its sequence IDs (as 
> returned from _changes) from strings to arrays. Is that true?
>
> I don’t have a local BigCouch to test against currently, but I just confirmed 
> that Cloudant is still using strings for sequence IDs; they look like this:
>> "last_seq":"5865-g1AAAADzeJzLYWBgYMlgTmFQSElKzi9KdUhJMtbLTS3KLElMT9VLzskvTUnMK9HLSy3JAapkSmRIsv___39WEgMD61ZUbeZ4tCU5AMmkepjOw6g6TfDozGMBkgwNQAqoeT9Y913i7YXoPgDRDbF7ShYAyj1QSA”
>
> If BigCouch is now using arrays, then I need to fix my code to serialize them 
> properly in the “?since” parameter to the _changes feed URL. But this is a 
> bit confusing — if I JSON-encode the sequence ID, then IDs that are strings 
> will now have double-quotes around them, and I don’t know if Cloudant will 
> understand that. Do I need to special-case based on the type, i.e. stripping 
> the quotes if it’s a string?
>
> —Jens

Reply via email to