Follow up improvements to Date_Time_Formatter

[radeusgd]

Pull Request Description

• Closes Additional improvements to date-time format #7872
• Also closes Make Data_Formatter.parse use Value_Type #7866

Important Notes

Checklist

Please ensure that the following checklist has been satisfied before submitting the PR:

• The documentation has been updated, if necessary.
• Screenshots/screencasts have been attached, if there are any visual changes. For interactive or animated visual changes, a screencast is preferred.
• All code follows the
  Scala,
  Java,
  and
  Rust
  style guides. In case you are using a language not listed above, follow the Rust style guide.
• All code has been tested:
  • Unit tests have been written where possible.
  • If GUI codebase was changed, the GUI was tested when built using ./run ide build.

[jdunkerley]

one slight question but looks good.

[radeusgd]

Added a benchmark (to Exploratory Benchmarks, so it will not be run by CI by default, I don't think it's important enough to run it, but IMO worth keeping in repo).

Raw results

Found 4 cases to execute
Benchmarking 'Format_Vector_Of_Dates.Naive' with configuration: [warmup={2 iterations, 5 seconds each}, measurement={2 iterations, 5 seconds each}]
Warmup duration:    11752.7002 ms
Warmup invocations: 3
Warmup avg time:    3910.385 ms
Measurement duration:    12369.5478 ms
Measurement invocations: 5
Measurement avg time:    2473.712 ms
Benchmark 'Format_Vector_Of_Dates.Naive' finished in 24129.273 ms
Benchmarking 'Format_Vector_Of_Dates.Prepared_Formatter' with configuration: [warmup={2 iterations, 5 seconds each}, measurement={2 iterations, 5 seconds each}]
Warmup duration:    10190.8434 ms
Warmup invocations: 1681
Warmup avg time:    5.949 ms
Measurement duration:    10242.1637 ms
Measurement invocations: 2431
Measurement avg time:    4.115 ms
Benchmark 'Format_Vector_Of_Dates.Prepared_Formatter' finished in 20434.452 ms
Benchmarking 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' with configuration: [warmup={2 iterations, 5 seconds each}, measurement={2 iterations, 5 seconds each}]
Warmup duration:    10040.4148 ms
Warmup invocations: 793
Warmup avg time:    12.621 ms
Measurement duration:    10045.1645 ms
Measurement invocations: 930
Measurement avg time:    10.761 ms
Benchmark 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' finished in 20086.788 ms
Benchmarking 'Format_Vector_Of_Dates.Column_Format' with configuration: [warmup={2 iterations, 5 seconds each}, measurement={2 iterations, 5 seconds each}]
Warmup duration:    10026.1589 ms
Warmup invocations: 549
Warmup avg time:    18.221 ms
Measurement duration:    10044.7375 ms
Measurement invocations: 637
Measurement avg time:    15.713 ms
Benchmark 'Format_Vector_Of_Dates.Column_Format' finished in 20072.428 ms

Re-creating the formatter at each invocation (default if you call something like vector.map _.format "str") is about 600x slower (!) than using a prepared formatter instance.

In comparison, using Column.format (which creates a formatter under the hood), is only ~4-5 times slower.

We should indeed try adding an LRU cache for the formats.

Results with shorter warmup (more CI ready)

Found 4 cases to execute
Benchmarking 'Format_Vector_Of_Dates.Naive' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    6927.1965 ms
Warmup invocations: 1
Warmup avg time:    6910.172 ms
Measurement duration:    8217.4258 ms
Measurement invocations: 3
Measurement avg time:    2738.048 ms
Benchmark 'Format_Vector_Of_Dates.Naive' finished in 15151.292 ms
Benchmarking 'Format_Vector_Of_Dates.Prepared_Formatter' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4048.7626 ms
Warmup invocations: 358
Warmup avg time:    11.188 ms
Measurement duration:    8127.921 ms
Measurement invocations: 969
Measurement avg time:    8.264 ms
Benchmark 'Format_Vector_Of_Dates.Prepared_Formatter' finished in 12178.561 ms
Benchmarking 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4008.1436 ms
Warmup invocations: 114
Warmup avg time:    35.117 ms
Measurement duration:    8023.6116 ms
Measurement invocations: 562
Measurement avg time:    14.244 ms
Benchmark 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' finished in 12033.032 ms
Benchmarking 'Format_Vector_Of_Dates.Column_Format' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4016.2714 ms
Warmup invocations: 172
Warmup avg time:    23.333 ms
Measurement duration:    8015.4114 ms
Measurement invocations: 413
Measurement avg time:    19.373 ms
Benchmark 'Format_Vector_Of_Dates.Column_Format' finished in 12033.088 ms

[radeusgd]

Benchmarks after implementing the cache

Raw results

Found 4 cases to execute
Benchmarking 'Format_Vector_Of_Dates.Naive' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4051.5825 ms
Warmup invocations: 372
Warmup avg time:    10.753 ms
Measurement duration:    8104.8076 ms
Measurement invocations: 1285
Measurement avg time:    6.229 ms
Benchmark 'Format_Vector_Of_Dates.Naive' finished in 12166.056 ms
Benchmarking 'Format_Vector_Of_Dates.Prepared_Formatter' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4051.4442 ms
Warmup invocations: 831
Warmup avg time:    4.814 ms
Measurement duration:    8175.2936 ms
Measurement invocations: 2022
Measurement avg time:    3.957 ms
Benchmark 'Format_Vector_Of_Dates.Prepared_Formatter' finished in 12228.495 ms
Benchmarking 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4008.8002 ms
Warmup invocations: 259
Warmup avg time:    15.449 ms
Measurement duration:    8026.8088 ms
Measurement invocations: 682
Measurement avg time:    11.734 ms
Benchmark 'Format_Vector_Of_Dates.Java_Formatter_recreated_on_each' finished in 12036.888 ms
Benchmarking 'Format_Vector_Of_Dates.Column_Format' with configuration: [warmup={1 iterations, 4 seconds each}, measurement={2 iterations, 4 seconds each}]
Warmup duration:    4013.7933 ms
Warmup invocations: 214
Warmup avg time:    18.735 ms
Measurement duration:    8061.3893 ms
Measurement invocations: 832
Measurement avg time:    9.617 ms
Benchmark 'Format_Vector_Of_Dates.Column_Format' finished in 12076.766 ms

Approach	Avg Measurement Time
Naïve (without cache)	2738.048 ms
Naïve (with cache)	6.229 ms
Formatter precomputed	3.957 ms
Java ofPattern re-created on each call	11.734 ms
Column.format	9.617 ms

Adding the cache was definitely worth it - on this example we are about 400x faster than before!

After implementing the cache all timings are on par between ~4-11ms.