Minor optimizations in hot codepaths accessing class parameters

[philippjfr]

No reason why the have to call .objects to get the list of parameters and also no reason we have check all parameters.

[philippjfr]

Actually this probably does not take into account dynamically added parameters added using .add_parameter

[philippjfr]

Should be fixed now.

[philippjfr]

Not showing a huge impact on the benchmarks but those don't cover real usecases yet, which end up calling __contains__ and a few other similar methods a whole lot.

       before           after         ratio
     [a654e784]       [e47cf4bd]
     <main>           <optimize_contains>
       41.9±0.5ms       41.1±0.3ms     0.98  benchmarks.ImportSuite.timeraw_import_param
      16.2±0.03μs      16.2±0.08μs     1.00  benchmarks.ParameterSuite.time_instantiation
      11.8±0.09μs      10.9±0.06μs     0.92  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_10_parameters_separate_cb
      11.8±0.08μs      10.8±0.02μs     0.92  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_10_parameters_shared_cb
      9.66±0.07μs      8.89±0.05μs     0.92  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_1_parameter
          423±2μs          412±6μs     0.97  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_10_parameters_separate_cb
          369±4μs          360±2μs     0.98  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_10_parameters_shared_cb
       51.4±0.1μs       50.6±0.2μs     0.99  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_1_parameter
          679±4μs          670±2μs     0.99  benchmarks.ParameterizedDependsSuite.time_declarative_10_parameters_separate_cb
          634±1μs          630±4μs     1.00  benchmarks.ParameterizedDependsSuite.time_declarative_10_parameters_shared_cb
          109±1μs        108±0.5μs     0.99  benchmarks.ParameterizedDependsSuite.time_declarative_1_parameter
          677±4μs          669±3μs     0.99  benchmarks.ParameterizedDependsSuite.time_watch_10_parameters_separate_cb
          640±4μs          626±1μs     0.98  benchmarks.ParameterizedDependsSuite.time_watch_10_parameters_shared_cb
          111±2μs        107±0.2μs     0.97  benchmarks.ParameterizedDependsSuite.time_watch_1_parameter
       31.9±0.3μs       30.1±0.1μs     0.94  benchmarks.ParameterizedInstantiateSuite.time_100_parameters
       11.8±0.1μs      10.8±0.06μs     0.92  benchmarks.ParameterizedInstantiateSuite.time_10_parameters
      9.79±0.08μs       9.06±0.3μs     0.93  benchmarks.ParameterizedInstantiateSuite.time_1_parameters
      3.10±0.02μs         3.08±0μs     1.00  benchmarks.ParameterizedSetattrSuite.time_class
      2.87±0.02μs      2.88±0.01μs     1.00  benchmarks.ParameterizedSetattrSuite.time_instance
       35.0±0.6μs       35.0±0.3μs     1.00  benchmarks.ParameterizedSuite.time_class_bare
      5.46±0.01ms      5.46±0.01ms     1.00  benchmarks.ParameterizedSuite.time_class_with_100_parameter
          583±3μs         596±20μs     1.02  benchmarks.ParameterizedSuite.time_class_with_10_parameter
       92.4±0.5μs         95.4±5μs     1.03  benchmarks.ParameterizedSuite.time_class_with_1_parameter
       12.5±0.2μs      12.4±0.04μs     0.99  benchmarks.WatcherSuite.time_trigger

[jbednar]

Looks good so far!

[philippjfr]

I'm done here for now, those py3.12 test failures seem unrelated.

[hoxbro]

The failing test is because of pytest-asyncio. The next release (0.23.3) should fix it.

pytest-dev/pytest-asyncio#711

[maximlt]

I added two simple benchmarks for x in <inst|class>.param(ParameterizedParamContainsSuite) which indeed show some significant speed-up. @philippjfr don't hesitate to add more "real-world" benchmarks if you have some in mind.

       before           after         ratio
     [a654e784]       [e47cf4bd]
     <main>           <optimize_contains>
       34.7±0.1ms       34.9±0.6ms     1.01  benchmarks.ImportSuite.timeraw_import_param
       16.8±0.1μs       16.5±0.1μs     0.98  benchmarks.ParameterSuite.time_instantiation
      8.81±0.05μs       8.16±0.1μs     0.93  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_10_parameters_separate_cb
      8.81±0.09μs      8.08±0.03μs     0.92  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_10_parameters_shared_cb
      7.03±0.03μs      6.51±0.02μs     0.93  benchmarks.ParameterizedDependsInstantiateSuite.time_declarative_1_parameter
          351±4μs          337±5μs     0.96  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_10_parameters_separate_cb
        313±0.5μs          314±8μs     1.00  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_10_parameters_shared_cb
       41.1±0.1μs       40.2±0.3μs     0.98  benchmarks.ParameterizedDependsInstantiateSuite.time_watch_1_parameter
          562±3μs         559±10μs     0.99  benchmarks.ParameterizedDependsSuite.time_declarative_10_parameters_separate_cb
        530±0.7μs          530±8μs     1.00  benchmarks.ParameterizedDependsSuite.time_declarative_10_parameters_shared_cb
       92.2±0.4μs       90.8±0.4μs     0.98  benchmarks.ParameterizedDependsSuite.time_declarative_1_parameter
          569±3μs         555±10μs     0.97  benchmarks.ParameterizedDependsSuite.time_watch_10_parameters_separate_cb
          530±3μs          527±2μs     1.00  benchmarks.ParameterizedDependsSuite.time_watch_10_parameters_shared_cb
       91.7±0.8μs         91.2±2μs     0.99  benchmarks.ParameterizedDependsSuite.time_watch_1_parameter
       25.7±0.1μs      24.5±0.07μs     0.95  benchmarks.ParameterizedInstantiateSuite.time_100_parameters
       8.93±0.2μs       8.30±0.2μs     0.93  benchmarks.ParameterizedInstantiateSuite.time_10_parameters
      7.06±0.09μs       6.48±0.1μs     0.92  benchmarks.ParameterizedInstantiateSuite.time_1_parameters
       65.9±0.6ns       67.1±0.7ns     1.02  benchmarks.ParameterizedParamAccessSuite.time_class
          175±1ns          177±1ns     1.01  benchmarks.ParameterizedParamAccessSuite.time_instance
-         502±2ns          174±2ns     0.35  benchmarks.ParameterizedParamContainsSuite.time_class
-         641±7ns          284±1ns     0.44  benchmarks.ParameterizedParamContainsSuite.time_instance
      2.61±0.02μs      2.61±0.01μs     1.00  benchmarks.ParameterizedSetattrSuite.time_class
      2.31±0.01μs      2.31±0.01μs     1.00  benchmarks.ParameterizedSetattrSuite.time_instance
       31.2±0.2μs       31.2±0.1μs     1.00  benchmarks.ParameterizedSuite.time_class_bare
      4.55±0.02ms      4.58±0.01ms     1.01  benchmarks.ParameterizedSuite.time_class_with_100_parameter
          492±4μs          490±3μs     0.99  benchmarks.ParameterizedSuite.time_class_with_10_parameter
       79.2±0.2μs       79.5±0.4μs     1.00  benchmarks.ParameterizedSuite.time_class_with_1_parameter
      8.63±0.09μs      8.61±0.07μs     1.00  benchmarks.WatcherSuite.time_trigger