Initialize the plugin by creating a new Delegate
delegate, err := ocr2keepers.NewDelegate(delegateConfig)Unit testing is used extensively and the primary goal is to keep test coverage above 70%.
Test coverage should be rerun with every commit either by running a git hook or make coverage to help maintain a high level of test coverage.
It is recommended that you install the git hooks so that the automated tooling is part of your workflow. Simply run:
cp -r .githooks/ .git/hooks
Explore test coverage per file with
$ go tool cover -html=cover.out
Benchmarking helps identify general function inefficiencies both with memory and processor time. Only benchmark functions that are likely to run multiple times, asynchronously, or be processor/memory intensive.
Using benchmarking consistently requires that os, arch, and cpu be kept consistent. Do not overwrite the benchmark.txt file unless your specs are identical.
To run benchmarking:
$ make benchmark | new.txt
To view a diff in benchmarks:
$ go install golang.org/x/perf/cmd/benchstat@latest
$ benchstat benchmarks.txt new.txt
To reduce dependencies on the main plugin repo, all loggers are based on the default go log.Logger. When using the NewDelegate function, a new logger is created with [keepers-plugin] prepending all logs and includes short file names/numbers. This logger writes its output to the ocr logger provided to the delegate as Debug logs.
The strategy of logging in this repo is to have two types of outcomes from logs:
- actionable - errors and panics (which should be handled by the plugin node itself)
- debug info - extra log info about inner workings of the plugin (optional based on provided ocr logger settings)
If an error cannot be handled, it should be bubbled up. If it cannot be bubbled up, it should panic. The plugin shouldn't be concerned with managing runtime errors, log severity, or panic recovery unless it cannot be handled by the plugin node process. An example might be a background service that is created a plugin startup but not managed by the plugin node. If there is such a service, it should handle its own recovery within the context of a Start/Stop service.