support fill(null)

[blalor]

It would be useful if fill could also support taking null as a value, so that a client would know that data is missing.

[pauldix]

I don't think there's a way to represent this in the JSON result. The null is really just the absence of a timestamped value. It's up to you charting library to represent that instead of just drawing a line between two other points. The best thing to do is to use bars or stacked bars and you'll see what you're expecting. There are other reasons stacked bars are superior: http://www.leancrew.com/all-this/2011/11/i-hate-stacked-area-charts/

I don't think there's anything actionable for us so I'm closing this out, but let me know if you have a use case I'm not thinking of.

[blalor]

You could just send null. :-)

Given a query like select mean(value) from "canaryio.state.ok.events.rate" group by time(10s) order asc, I'll get the following back:

[{u'columns': [u'time', u'mean'],
  u'name': u'canaryio.state.ok.events.rate',
  u'points': [[1404378370, 11.9],
              [1404378380, 11.7],
              [1404378420, 5.6],
              [1404378430, 12.4],
              [1404378440, 12.9]]}]

The points at 1404378390, 1404378400 and 1404378410 are missing. If I add fill(0) I'll get:

[{u'columns': [u'time', u'mean'],
  u'name': u'canaryio.state.ok.events.rate',
  u'points': [[1404378370, 11.9],
              [1404378380, 11.7],
              [1404378390, 0],
              [1404378400, 0],
              [1404378410, 0],
              [1404378420, 5.6],
              [1404378430, 12.4],
              [1404378440, 12.9]]}]

What I want is to use fill(null) to get:

[{u'columns': [u'time', u'mean'],
  u'name': u'canaryio.state.ok.events.rate',
  u'points': [[1404378370, 11.9],
              [1404378380, 11.7],
              [1404378390, null],
              [1404378400, null],
              [1404378410, null],
              [1404378420, 5.6],
              [1404378430, 12.4],
              [1404378440, 12.9]]}]

This is what Graphite does for missing buckets, and it makes it so that the client doesn't have to do missing-value detection on its own. You're already supporting fill(0), so it seems like a logical extension to also support null.

[pauldix]

I guess you're right :)

[Dieterbe]

+1 this will also be nice for the graphite-influxdb backend. Presumably influx can fill gaps faster than a retroactive (python) loop. + no need to reimplement the same thing in multiple clients. Thanks!