New benchmarks for parsing dates with dayfirst=True and format='%d-%m-%Y'

[anmyachev]

• closes read_csv : using day first 23x to 35x slower than setting the format explicitly #25848
• tests added
• passes git diff upstream/master -u -- "*.py" | flake8 --diff

Gap in performance between use dayfirst or format is acceptable now.

asv run -E existing -b ParseDateComparison -a warmup_time=2 -a sample_time=2:

io.csv.ParseDateComparison.time_read_csv_dayfirst	parameter	time
-	cache_dates	-
-	False	16.3±0.4ms
-	True	7.22±0.2ms
io.csv.ParseDateComparison.time_to_datetime_dayfirst	parameter	time
-	cache_dates	-
-	False	15.0±0.3ms
-	True	5.67±0.2ms
io.csv.ParseDateComparison.time_to_datetime_format_DD_MM_YYYY	parameter	time
-	cache_dates	-
-	False	25.0±0.4ms
-	True	5.47±0.4ms

[jreback]

great, can you show results for thee in the top-section.

[codecov]

Codecov Report

Merging #26360 into master will decrease coverage by 0.13%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #26360      +/-   ##
==========================================
- Coverage   91.81%   91.68%   -0.14%     
==========================================
  Files         175      174       -1     
  Lines       52289    50700    -1589     
==========================================
- Hits        48009    46482    -1527     
+ Misses       4280     4218      -62

Flag	Coverage Δ
#multiple	90.18% <ø> (-0.18%)	⬇️
#single	41.19% <ø> (+0.32%)	⬆️

Impacted Files	Coverage Δ
pandas/io/gbq.py	78.94% <0%> (-10.53%)	⬇️
pandas/core/frame.py	97.01% <0%> (-0.12%)	⬇️
pandas/io/parsers.py

---

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 4a30fa5...3b1034b. Read the comment docs.

[codecov]

Codecov Report

Merging #26360 into master will decrease coverage by 0.12%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #26360      +/-   ##
==========================================
- Coverage   91.81%   91.68%   -0.13%     
==========================================
  Files         175      174       -1     
  Lines       52289    50749    -1540     
==========================================
- Hits        48009    46530    -1479     
+ Misses       4280     4219      -61

Flag	Coverage Δ
#multiple	90.19% <ø> (-0.17%)	⬇️
#single	41.16% <ø> (+0.29%)	⬆️

Impacted Files	Coverage Δ
pandas/io/gbq.py	78.94% <0%> (-10.53%)	⬇️
pandas/compat/numpy/function.py	90.39% <0%> (-0.41%)	⬇️
pandas/core/frame.py	97.02% <0%> (-0.12%)	⬇️
pandas/core/series.py	93.67% <0%> (ø)	⬆️
pandas/io/parsers.py
pandas/core/indexes/base.py	96.72% <0%> (ø)	⬆️
pandas/core/arrays/integer.py	96.35% <0%> (+0.02%)	⬆️
pandas/core/sparse/frame.py	95.63% <0%> (+0.14%)	⬆️
pandas/core/arrays/sparse.py	92.7% <0%> (+0.39%)	⬆️

---

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 4a30fa5...e33284d. Read the comment docs.

[jreback]

@anmyachev

@jreback Should I make the conditions for testing the same? (for this, most likely I will have to create new classes)

yes making these consistent for benchmarks (and testing both cache on/off would be great)

[anmyachev]

yes making these consistent for benchmarks (and testing both cache on/off would be great)

@jreback this is done.

[jreback]

thanks @anmyachev