REF: avoid getattr pattern for join_indexer

[jbrockmendel]

No description provided.

[jreback]

can u perf check to make sure that does the right thing

[jbrockmendel]

can u perf check to make sure that does the right thing

Noise

    before     after       ratio
  [e623f0fb] [7b88515e]
+  135.77ms   292.28ms      2.15  binary_ops.Ops.time_frame_multi_and(True, 'default')
+   13.20ms    19.18ms      1.45  binary_ops.Ops.time_frame_add(True, 'default')
+    2.10ms     2.74ms      1.31  stat_ops.Correlation.time_corr('pearson', True)
+    7.23μs     8.76μs      1.21  timeseries.DatetimeIndex.time_get('tz_aware')
+   13.04μs    15.67μs      1.20  offset.OffestDatetimeArithmetic.time_apply(<YearEnd: month=12>)
+    1.92μs     2.30μs      1.20  timestamp.TimestampOps.time_tz_convert(<UTC>)
+   27.29μs    32.50μs      1.19  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<Day>)
+   13.31ms    15.82ms      1.19  timeseries.ToDatetimeFormat.time_no_exact
+   14.45μs    17.10μs      1.18  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<BusinessQuarterEnd: startingMonth=3>)
+   13.34μs    15.63μs      1.17  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<MonthEnd>)
+   13.09ms    15.31ms      1.17  rolling.EWMMethods.time_ewm('DataFrame', 1000, 'float', 'std')
+   56.09ms    65.07ms      1.16  eval.Eval.time_add('python', 'all')
+    1.28ms     1.48ms      1.15  inference.NumericInferOps.time_subtract(<class 'numpy.int32'>)
+    4.94ms     5.70ms      1.15  strings.Cat.time_cat(0, ',', '-', 0.0)
+    2.34μs     2.69μs      1.15  timestamp.TimestampOps.time_replace_None(tzutc())
+   15.80ms    18.16ms      1.15  gil.ParallelReadCSV.time_read_csv('object')
+    6.36μs     7.27μs      1.14  timestamp.TimestampOps.time_normalize(None)
+    3.68μs     4.19μs      1.14  timedelta.TimedeltaConstructor.time_from_np_timedelta
+  564.77μs   642.94μs      1.14  stat_ops.Correlation.time_corr_series('kendall', True)
+    4.94ms     5.63ms      1.14  inference.DateInferOps.time_subtract_datetimes
+    6.55μs     7.44μs      1.14  timeseries.AsOf.time_asof_single_early('Series')
+   17.61ms    19.99ms      1.13  frame_methods.MaskBool.time_frame_mask_bools
+    2.54ms     2.88ms      1.13  inference.NumericInferOps.time_divide(<class 'numpy.int64'>)
+   13.01μs    14.73μs      1.13  offset.OffestDatetimeArithmetic.time_apply(<MonthEnd>)
+     1.19s      1.35s      1.13  timeseries.Iteration.time_iter(<function period_range at 0x7f11b4b42b70>)
+   17.21μs    19.45μs      1.13  period.PeriodUnaryMethods.time_now('M')
+  125.99ms   142.15ms      1.13  timeseries.ToDatetimeCache.time_dup_string_tzoffset_dates(False)
+   13.72μs    15.45μs      1.13  offset.OffestDatetimeArithmetic.time_apply(<BusinessMonthBegin>)
+   23.05ms    25.93ms      1.13  strings.Methods.time_title
+  265.95ms   298.69ms      1.12  io.json.ToJSON.time_to_json('index', 'df_int_floats')
+     8.99s     10.09s      1.12  stat_ops.Correlation.time_corr_wide('kendall', False)
+  901.09ns     1.01μs      1.12  period.PeriodProperties.time_property('min', 'dayofweek')
+   23.52μs    26.34μs      1.12  offset.OffestDatetimeArithmetic.time_subtract_10(<BusinessQuarterEnd: startingMonth=3>)
+   34.80ms    38.95ms      1.12  gil.ParallelGroupbyMethods.time_parallel(2, 'prod')
+  819.06μs   915.96μs      1.12  stat_ops.Correlation.time_corr_series('spearman', True)
+  572.74μs   640.07μs      1.12  stat_ops.Correlation.time_corr_series('kendall', False)
+   10.10ms    11.28ms      1.12  algorithms.Duplicated.time_duplicated('first', 'string')
+    3.50μs     3.91μs      1.11  indexing_engines.ObjectEngineIndexing.time_get_loc('monotonic_incr')
+   14.17μs    15.78μs      1.11  offset.OffestDatetimeArithmetic.time_apply(<SemiMonthEnd: day_of_month=15>)
+   23.55μs    26.20μs      1.11  offset.OffestDatetimeArithmetic.time_add(<CustomBusinessDay>)
+  135.84ms   151.01ms      1.11  inference.ToNumericDowncast.time_downcast('string-nint', None)
+    7.64ms     8.48ms      1.11  series_methods.NanOps.time_func('kurt', 1000000, 'int64')
+   27.82μs    30.85μs      1.11  timestamp.TimestampProperties.time_is_quarter_end(<DstTzInfo 'Europe/Amsterdam' LMT+0:20:00 STD>, 'B')
+  390.29ms   432.65ms      1.11  rolling.Apply.time_rolling('Series', 1000, 'int', <function sum at 0x7f11bbdd99d8>, True)
+  891.93ns   988.13ns      1.11  period.PeriodProperties.time_property('M', 'month')
+   13.77μs    15.26μs      1.11  series_methods.SearchSorted.time_searchsorted('int32')
+    8.58ms     9.49ms      1.11  timeseries.Factorize.time_factorize('Asia/Tokyo')
+   12.81μs    14.16μs      1.11  offset.OffestDatetimeArithmetic.time_apply(<MonthBegin>)
+   25.46ms    28.13ms      1.11  io.sql.WriteSQLDtypes.time_to_sql_dataframe_column('sqlite', 'bool')
+   14.45μs    15.96μs      1.10  offset.OffestDatetimeArithmetic.time_add(<BusinessMonthBegin>)
+     9.15s     10.10s      1.10  stat_ops.Correlation.time_corr_wide('kendall', True)
+   21.89μs    24.15μs      1.10  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<CustomBusinessDay>)
+   14.26μs    15.73μs      1.10  offset.OffestDatetimeArithmetic.time_add(<MonthBegin>)
+  445.02ms   490.51ms      1.10  rolling.Apply.time_rolling('Series', 1000, 'int', <function Apply.<lambda> at 0x7f11d5229598>, True)
+   24.81μs    27.33μs      1.10  offset.OffestDatetimeArithmetic.time_subtract_10(<SemiMonthEnd: day_of_month=15>)
+   60.44ms    66.56ms      1.10  gil.ParallelFactorize.time_parallel(4)
-    7.73μs     7.02μs      0.91  indexing.NonNumericSeriesIndexing.time_getitem_scalar('string', 'unique_monotonic_inc')
-    2.12ms     1.92ms      0.91  indexing.NumericSeriesIndexing.time_loc_scalar(<class 'pandas.core.indexes.numeric.UInt64Index'>, 'nonunique_monotonic_inc')
-   17.03ms    15.44ms      0.91  frame_methods.MaskBool.time_frame_mask_floats
-   75.17ms    68.04ms      0.91  gil.ParallelGroupbyMethods.time_parallel(4, 'prod')
-   14.77μs    13.37μs      0.91  offset.OffestDatetimeArithmetic.time_apply(<BusinessQuarterEnd: startingMonth=3>)
-   30.46μs    27.48μs      0.90  series_methods.All.time_all(1000, 'slow')
-  216.05ns   194.88ns      0.90  timestamp.TimestampProperties.time_weekday_name(<DstTzInfo 'Europe/Amsterdam' LMT+0:20:00 STD>, None)
-    5.75ms     5.19ms      0.90  gil.ParallelRolling.time_rolling('mean')
-   10.30μs     9.28μs      0.90  indexing.NumericSeriesIndexing.time_getitem_scalar(<class 'pandas.core.indexes.numeric.Int64Index'>, 'nonunique_monotonic_inc')
-   16.47μs    14.81μs      0.90  offset.OffestDatetimeArithmetic.time_add(<BusinessMonthEnd>)
-    8.27ms     7.43ms      0.90  io.csv.ParseDateComparison.time_to_datetime_dayfirst(False)
-    3.70μs     3.33μs      0.90  timeseries.DatetimeIndex.time_get('dst')
-  239.93ms   215.40ms      0.90  multiindex_object.Values.time_datetime_level_values_copy
-   15.54μs    13.92μs      0.90  series_methods.SearchSorted.time_searchsorted('uint32')
-   14.77μs    13.23μs      0.90  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<QuarterEnd: startingMonth=3>)
-   76.47ms    68.49ms      0.90  strings.Repeat.time_repeat('array')
-   48.40ms    43.35ms      0.90  timedelta.ToTimedeltaErrors.time_convert('coerce')
-   15.80μs    14.12μs      0.89  offset.OffestDatetimeArithmetic.time_apply_np_dt64(<YearBegin: month=1>)
-    1.87μs     1.67μs      0.89  timestamp.TimestampConstruction.time_parse_today
-   11.00μs     9.82μs      0.89  timestamp.TimestampProperties.time_is_quarter_start(tzutc(), 'B')
-   48.04ms    42.87ms      0.89  io.parsers.DoesStringLookLikeDatetime.time_check_datetimes('2Q2005')
-    7.10μs     6.33μs      0.89  timestamp.TimestampOps.time_normalize(<UTC>)
-   51.82μs    46.18μs      0.89  multiindex_object.GetLoc.time_med_get_loc
-  577.90μs   514.94μs      0.89  multiindex_object.Duplicates.time_remove_unused_levels
-   17.18μs    15.25μs      0.89  categoricals.Contains.time_categorical_contains
-   15.23μs    13.52μs      0.89  offset.OffestDatetimeArithmetic.time_apply(<BusinessMonthEnd>)
-    4.91μs     4.33μs      0.88  timedelta.TimedeltaConstructor.time_from_int
-    5.81ms     5.10ms      0.88  gil.ParallelRolling.time_rolling('var')
-   12.65μs    11.08μs      0.88  timestamp.TimestampOps.time_to_julian_date('US/Eastern')
-  105.53ms    92.30ms      0.87  sparse.SparseSeriesToFrame.time_series_to_frame
-    2.46ms     2.14ms      0.87  categoricals.Concat.time_union
-   23.74μs    20.43μs      0.86  offset.OffestDatetimeArithmetic.time_subtract(<QuarterBegin: startingMonth=3>)
-   23.19μs    19.91μs      0.86  period.PeriodUnaryMethods.time_asfreq('min')
-   25.57ms    21.88ms      0.86  stat_ops.FrameOps.time_op('median', 'float', 1, False)
-  185.16μs   157.25μs      0.85  period.PeriodConstructor.time_period_constructor('D', False)
-   27.40μs    23.22μs      0.85  offset.OffestDatetimeArithmetic.time_subtract_10(<QuarterEnd: startingMonth=3>)
-   14.28μs    12.06μs      0.84  timestamp.TimestampOps.time_replace_None('US/Eastern')
-   28.62μs    23.97μs      0.84  timestamp.TimestampProperties.time_month_name(<DstTzInfo 'Europe/Amsterdam' LMT+0:20:00 STD>, None)
-  290.44ns   236.05ns      0.81  timestamp.TimestampProperties.time_is_month_start(tzutc(), None)
-    3.27ms     2.62ms      0.80  binary_ops.Ops2.time_frame_series_dot
-   25.24μs    20.09μs      0.80  offset.OffestDatetimeArithmetic.time_subtract_10(<BusinessMonthEnd>)
-    2.59ms     2.02ms      0.78  indexing.IntervalIndexing.time_getitem_scalar
-   18.03μs    13.95μs      0.77  series_methods.SearchSorted.time_searchsorted('int8')
-    1.16μs   900.98ns      0.77  period.PeriodProperties.time_property('M', 'year')
-  205.07ms   153.54ms      0.75  binary_ops.Ops2.time_frame_dot
-   19.81μs     7.74μs      0.39  algorithms.MaybeConvertObjects.time_maybe_convert_objects

SOME BENCHMARKS HAVE CHANGED SIGNIFICANTLY.

[WillAyd]

lgtm @jbrockmendel a lot of nice cleanups lately great job

[jreback]

at some point if can change to use a fixture to ensure more coverage (we may need to expand the indexers here as well), pls create a followup issue.

[jreback]

lgtm, followup issue.