Do not allocate MessageIDs on the heap

[dferstay]

Passing a function parameter by pointer (or writing a pointer into a
channel) will cause the Go escape analysis to allocate the parameter on
the heap.

This change passes messageID struct instances by value instead of by
pointer; this keeps messageID structs on the stack.

Each message produced or consumed by the library is associated with
a MessageID; keeping instances of the MessageID structure on the stack
reduces heap allocation and associated GC cost.

Signed-off-by: Daniel Ferstay [email protected]

Motivation

Reduce the amount of per-Message heap allocation performed by the library

Modifications

This change modifies the Consumer and Producer code paths to pass messageID struct instances by value instead of by
pointer; this keeps the messageID structures on the stack and off of the heap.

Verifying this change

• Make sure that the change passes the CI checks.

This change is already covered by existing tests, such as:

• pulsar/consumer_multitopic_test.go
• pulsar/consumer_partition_test.go
• pulsar/consumer_regex_test.go
• pulsar/consumer_test.go
• pulsar/impl_message_test.go
• pulsar/negative_acks_tracker_test.go
• pulsar/producer_test.go
• pulsar/reader_test.go

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): no
• The public API: no
• The schema: no
• The default values of configurations: no
• The wire protocol: no

Documentation

• Does this pull request introduce a new feature? no

[dferstay]

Below is the alloc_space view of a pprof heap profile collected during a large scale test of a high-throughput service that uses pulsar-client-go to send messages to Pulsar. Note that messageID structures account for 6.07% of all allocation; this percentage grows the longer the service runs as messageIDs escape to the heap for every message produced.

[wolfstudy]

Cool, LGTM +1