Remove arg that is only ever used as NPY_UNSAFE_CASTING

[jbrockmendel]

Several functions from src/datetime are only ever called with the casting rule NPY_UNSAFE_CASTING. By getting rid of that dummy argument, the remaining code gets simplified quite a bit.

This PR removes that argument, then removes code that this renders unreachable or unused. It also removes several commented-out functions.

There are a couple of other never-used args; taking these one at a time.

[codecov]

Codecov Report

Merging #18546 into master will decrease coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #18546      +/-   ##
==========================================
- Coverage   91.35%   91.33%   -0.02%     
==========================================
  Files         164      164              
  Lines       49802    49802              
==========================================
- Hits        45496    45487       -9     
- Misses       4306     4315       +9

Flag	Coverage Δ
#multiple	89.13% <ø> (ø)	⬆️
#single	40.81% <ø> (-0.07%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	25% <0%> (-58.34%)	⬇️
pandas/core/frame.py	97.81% <0%> (-0.1%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2a0e54b...02e4e8b. Read the comment docs.

[codecov]

Codecov Report

Merging #18546 into master will decrease coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #18546      +/-   ##
==========================================
- Coverage   91.35%   91.33%   -0.02%     
==========================================
  Files         164      164              
  Lines       49802    49802              
==========================================
- Hits        45496    45487       -9     
- Misses       4306     4315       +9

Flag	Coverage Δ
#multiple	89.13% <ø> (ø)	⬆️
#single	40.81% <ø> (-0.07%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	25% <0%> (-58.34%)	⬇️
pandas/core/frame.py	97.81% <0%> (-0.1%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2a0e54b...02e4e8b. Read the comment docs.

[jreback]

can you run an asv set on timeseries and see if anything changes.

[jreback]

do we need an assertion here that unit is FR_ns?

[jbrockmendel]

To double-check, what line are you referring to? The answer is "no" regardless, but the reason why may vary.

[jreback]

parse_iso_8601_datetimetime takes a unit arg, is this vaidated that it is only ever FR_ns

[jbrockmendel]

OK. The answer is "no" and the reason is because 1) the unit arg is no longer used once we've gotten rid of the casting arg (so it'll be removed in a follow-up) and 2) parse_iso_8601_datetime is only ever called from src/datetime.pxd and in that one case it is with PANDAS_FR_ns.

[jreback]

ok that's what I figured, ok if you are removing then this is ok.

[jbrockmendel]

can you run an asv set on timeseries and see if anything changes.

Starting now.

[jbrockmendel]

taskset 2 asv continuous -f 1.1 -E virtualenv master HEAD -b timeseries
[...]
       before           after         ratio
     [2a0e54bc]       [02e4e8bf]
+      19.0±0.1ms            135ms     7.10  timeseries.DatetimeIndex.time_dti_tz_factorize
+        54.9±5ms         117±20ms     2.12  timeseries.DatetimeIndex.time_dti_factorize
+           1.61s            1.93s     1.20  timeseries.AsOfDataFrame.time_asof
-     2.49±0.01ms      2.20±0.02ms     0.88  timeseries.ToDatetime.time_cache_true_with_dup_string_dates
-      7.85±0.4μs      6.75±0.04μs     0.86  timeseries.DatetimeIndex.time_timestamp_tzinfo_cons
-      25.3±0.6μs       20.2±0.2μs     0.80  offset.Day.time_timeseries_day_incr
-           1.83s        1.12±0.2s     0.61  timeseries.AsOfDataFrame.time_asof_nan

asof and dti_factorize I've gotten use to completely ignoring. Re-running anyway, but this looks like noise.

[jbrockmendel]

       before           after         ratio
     [2a0e54bc]       [02e4e8bf]
+     7.81±0.07ms         23.8±3ms     3.05  timeseries.DatetimeIndex.time_dti_tz_factorize
+         389±2μs         444±20μs     1.14  timeseries.DatetimeIndex.time_reset_index
-         134±8μs        113±0.1μs     0.85  timeseries.DatetimeIndex.time_unique
-         162±2ms          137±2ms     0.85  timeseries.ToDatetime.time_iso8601_tz_spaceformat
-        27.6±2ms       22.4±0.2ms     0.81  timeseries.DatetimeIndex.time_to_pydatetime

[jbrockmendel]

      before           after         ratio
     [2a0e54bc]       [02e4e8bf]
+        62.1±2ms         101±30ms     1.63  timeseries.AsOfDataFrame.time_asof_nan
+     1.47±0.01ms      1.77±0.03ms     1.20  timeseries.ToDatetime.time_cache_false_with_dup_string_dates_and_format
+      13.3±0.4μs       15.2±0.1μs     1.14  offset.YearBegin.time_timeseries_year_incr
-      19.8±0.1μs       17.6±0.2μs     0.89  timeseries.AsOf.time_asof_single_early
-      3.96±0.2ms      3.44±0.03ms     0.87  timeseries.ToDatetime.time_cache_true_with_unique_seconds_and_unit
-      2.00±0.1ms         1.69±0ms     0.84  timeseries.DatetimeIndex.time_add_timedelta

[jbrockmendel]

There are a few of these; hoping to get them out of the way before doing away with datetime.pxd

[jreback]

thanks!