Parse statsd lines with multiple metric bits

[MerlinDMC]

This resembles the parsing which is done in etsy/statsd to add multiple datapoints for a measurement in a single line.
https://github.com/etsy/statsd/blob/243a1f2a166c2d573f7582dc0c42f50f257e4150/stats.js#L225-L277

It seems as if some client applications use those possibilities to pack more data into each UDP packet and simply omit the repeating measurement name where possible.

An example for this is the Kamon.io statsd reporter which produces lines like these:

kamon.host.akka-actor.kamon_user_kamon-system-metrics_sigar-metrics-recorder.processing-time:581632|ms:593920|ms:610304|ms:622592|ms:634880|ms:638976|ms:659456|ms:679936|ms:720896|ms:770048|ms
kamon.host.akka-actor.kamon_user_kamon-system-metrics_sigar-metrics-recorder.time-in-mailbox:5664|ms:6176|ms:6240|ms|@0.5:6304|ms:6336|ms:6656|ms:6720|ms:7072|ms:12288|ms

[sparrc]

@MerlinDMC Is this documented in the StatsD spec anywhere? I didn't realize that this was supported by the standard StatsD implementation.

You'll need to also add documentation for this to the Telegraf StatsD README

[sparrc]

Can you add more tests here? I'd like to see:

without sample rate:

"valid.multiple.nosample:0|ms|:1|ms:200|ms"

Duplicate measurements (also, document what would happen here):

"valid.multiple.duplicate:0|ms:1|ms:0|ms:1|ms"

Duplicate measurements and different types (documentation needed here too):

"valid.multiple.duplicate:1|c:1|c:2|c:1|c"
"valid.multiple.duplicate:1|h:1|h:2|h:1|h"
"valid.multiple.duplicate:1|s:1|s:2|s:1|s"
"valid.multiple.duplicate:1|g:1|g:2|g:1|g" <-- what happens here? gauge will be set to the final value?

Multiple measurement types w/ one bucket (is this supported?)

"valid.multiple.duplicate:1|c:1|ms:2|s:1|g"

[MerlinDMC]

@sparrc It doesn't seem to be documented in the spec. Also if "the spec" is https://github.com/b/statsd_spec - that seems to miss some other stuff as well - sample rate for timers for example which are in etsy/statsd since Feb 2013.

The overall behaviour seems to be that metric name and data packets are separated by a colon and each data packet needs at least a value and a type separated by a pipe.

foo:1|c

As soon as multiple data packets are on a single line the metric name is reused throughout them.

foo:1|c:+1|g

same as

foo:1|c
foo:+1|g

To test/prove that behaviour you can run https://github.com/MerlinDMC/statsd-multiple-metrics-prove which will start two etsy/statsd instances and feed them with the same data and show their aggregated output on the console.

• one with 1 stat per line
• one with multiple stats per line

I've change the added test to feed two statsd instances and ensure the results are the same as well - so feeding single stats and multiple will compute to the same values in the end.

I would want to squash this feature branch into a single commit once it's good for a merge if possible.

[sparrc]

@MerlinDMC I'm sorry for the late eyes on your latest changes, looks good to me, is it ready to merge?

[MerlinDMC]

@sparrc I did rebase and squash it on top of your current master branch. If the tests on circleci are good feel free to merge this.

[sparrc]

Thank you @MerlinDMC! I'll get this merged