Do not export empty span lists

[dyladan]

Fixes #895

[codecov-io]

Codecov Report

Merging #896 into master will decrease coverage by 0.37%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #896      +/-   ##
==========================================
- Coverage   94.19%   93.81%   -0.38%     
==========================================
  Files         248      223      -25     
  Lines       10847     8793    -2054     
  Branches     1059      857     -202     
==========================================
- Hits        10217     8249    -1968     
+ Misses        630      544      -86     

Impacted Files	Coverage Δ
...telemetry-tracing/src/export/BatchSpanProcessor.ts	95.74% <100.00%> (+0.09%)	⬆️
...try-tracing/test/export/BatchSpanProcessor.test.ts	100.00% <100.00%> (ø)
packages/opentelemetry-plugin-mysql/src/utils.ts	90.90% <0.00%> (-4.55%)	⬇️
packages/opentelemetry-resources/src/constants.ts	100.00% <0.00%> (ø)
packages/opentelemetry-plugin-mysql/src/version.ts	100.00% <0.00%> (ø)
...ckages/opentelemetry-plugin-mongodb/src/version.ts	100.00% <0.00%> (ø)
...kages/opentelemetry-node/test/registration.test.ts	100.00% <0.00%> (ø)
...ages/opentelemetry-resources/test/Resource.test.ts	100.00% <0.00%> (ø)
...opentelemetry-api/src/metrics/NoopMeterProvider.ts	100.00% <0.00%> (ø)
...lemetry-resources/test/resource-assertions.test.ts	100.00% <0.00%> (ø)
... and 37 more

[mayurkale22]

lgtm

[mayurkale22]

@pauldraper would be nice if you can review this one.

[pauldraper]

Looks good.

I do think that it would be even better to replace setInterval with setTimeout, and invoke it only when there is something to send.

In-memory polling is poor citizenry for an application. On dedicated hardware (e.g. server), it's fine, but for other cases (e.g. a desktop app), it's impolite to constantly be using the CPU and preventing memory from paging to disk.

[dyladan]

@pauldraper so instead of having a timer fire every x seconds no matter what, you would have the timer start when the first ended span is added to the list, then when the timer fires it exports all ended spans. There is then not another timer set until onEnd is called again?

[dyladan]

@pauldraper just updated to remove the in-memory polling. PTAL

[pauldraper]

LGTM

[dyladan]

Re-requesting reviews because of the changes suggested by @pauldraper

[mayurkale22]

Added one optional nit, otherwise logic lgtm.