Add cache for TableRecord.defaultDatabaseTableName

[gjeck]

Resolves #637. Reduces cpu/memory impact from repeated calls to TableRecord.defaultDatabaseTableName by cacheing the name in a private global NSCache.

Pull Request Checklist

• This pull request is submitted against the development branch.
• Inline documentation has been updated.
• README.md or another dedicated guide has been updated.
• Changes are tested.

[groue]

That's lovely, @gjeck 👍

Did you run some benchmarks and see any difference?

[gjeck]

I wasn't sure if it was worth committing but I did write a very basic performance test that potentially demonstrated the improvement. This was the test:

class TableRecordPerformanceTests: XCTestCase {
    func testPerformanceExample() {
        self.measure {
            (0...10000).forEach { _ in
                _ = MockRecord.databaseTableName
            }
        }
    }
}
private struct MockRecord: TableRecord {}

This was the result before adding the cache:

Test Case '-[GRDBOSXPerformanceTests.TableRecordPerformanceTests testPerformanceExample]' measured [Time, seconds] average: 0.148, relative standard deviation: 4.136%, values: [0.165622, 0.147685, 0.146469, 0.145622, 0.146661, 0.147595, 0.149581, 0.142303, 0.146900, 0.143601], performanceMetricID:com.apple.XCTPerformanceMetric_WallClockTime, baselineName: "", baselineAverage: , maxPercentRegression: 10.000%, maxPercentRelativeStandardDeviation: 10.000%, maxRegression: 0.100, maxStandardDeviation: 0.100

This was the result after adding the cache:

Test Case '-[GRDBOSXPerformanceTests.TableRecordPerformanceTests testPerformanceExample]' measured [Time, seconds] average: 0.022, relative standard deviation: 10.138%, values: [0.028496, 0.021501, 0.020826, 0.021478, 0.022193, 0.019994, 0.021537, 0.021261, 0.021149, 0.021637], performanceMetricID:com.apple.XCTPerformanceMetric_WallClockTime, baselineName: "", baselineAverage: , maxPercentRegression: 10.000%, maxPercentRelativeStandardDeviation: 10.000%, maxRegression: 0.100, maxStandardDeviation: 0.100

So not an earth shattering improvement, but the average case seems quite a bit better (0.148 vs 0.022).

[groue]

That's still an improvement!

[groue]

Welcome in, @gjeck 🤗 Would you like to be granted a push access to the repository? I'll invite you right away.

[gjeck]

Sweet! Yeah that would be awesome, thank you 😄

[groue]

Shipped in v4.9.0 :-)