Performance improvements in TemplateVariable.essence()

[MikeRocke]

Noticed this took a relatively large portion of time when Unit tests of LinkBuilding (via Integration testing and Testing Controller layer), due to String format needing to parse the format string for each template variable

Minor spell corrections on test methods when spotted

[odrotbohm]

Nice one! Do you have any numbers hand pre and post optimization?

[MikeRocke]

Hey there!

Most welcome

I have some rough numbers from last night.

First rough exercise, just to loop through many times for same input,

        String name = "make";
        int limit = 2;
        TemplateVariable.Cardinality cardinality = TemplateVariable.Cardinality.COMPOSITE;
        boolean composite = cardinality.equals(TemplateVariable.Cardinality.COMPOSITE);

        //510ms
        for (int i = 0; i < 100000; i++) {
            String originalValue = String.format("%s%s%s", name, limit != -1 ? ":".concat(String.valueOf(limit)) : "", composite ? "*" : "");
        }

and

        String name = "make";
        int limit = 2;
        TemplateVariable.Cardinality cardinality = TemplateVariable.Cardinality.COMPOSITE;
        boolean composite = cardinality.equals(TemplateVariable.Cardinality.COMPOSITE);

        //25ms
        for (int i = 0; i < 100000; i++) {
            StringBuilder stringBuilder = new StringBuilder(name);
            if (limit != -1) {
                stringBuilder = stringBuilder.append(":").append(String.valueOf(limit));
            }
            if (composite) {
                stringBuilder = stringBuilder.append("*");
            }
            String newvalue = stringBuilder.toString();
        }

Approx ~500ms before, and ~25ms afterwards

Unfortunately, don't have the evidence to hand on the next exercise, but I seen a ~33% improvement on my build duration, as they tend to perform requests to RestControllers with unique data for each test

And spotted this cause with async profiler in intellij, that the essence method was a big contributor

Appreciate these aren't the most scientific methods, and am happy to get some more conclusive before and afters

[odrotbohm]

Thanks. I took a few extra steps to remove all usage of String.format(…) in that path, added a quick benchmark calling ….toString() and got from this:

Benchmark                                    Mode  Cnt         Score        Error  Units
TemplateVariableBenchmark.toString(…)       thrpt    3   2803239,270 ± 110258,955  ops/s

to this:

Benchmark                                    Mode  Cnt         Score        Error  Units
TemplateVariableBenchmark.toString(…)       thrpt    3  10753653,459 ± 156684,459  ops/s

which is almost a 4x improvement. Pushes coming in a minute.

[MikeRocke]

Fantastic! Thank you!

[odrotbohm]

Backported until 1.4.x. Feel free to give the snapshots a try.

[MikeRocke]

Just an update, quick experiment of v1.4.1 against 1.4.2-SNAPSHOT in one of our app build, test phase went from 800s to 534s. I know I should repeat the experiments to get a good statistical figure but overall, looks like a win! Super thank you @odrotbohm