Fastest way to count the number of features in a feature class?

Question

With the introduction of the Data Access module in arcpy (30x faster search cursors), I want to know if counting features matching sql criteria is faster than the traditional MakeTableView + GetCount methodology?

Answer 1

I am using an example with 1 million randomly generated points inside of a filegeodatabase. Attached here.

Here is some code to get us started:

import time
import arcpy

arcpy.env.workspace = "C:\CountTest.gdb"

time.sleep(5) # Let the cpu/ram calm before proceeding!

"""Method 1"""
StartTime = time.clock()
with arcpy.da.SearchCursor("RandomPoints", ["OBJECTID"]) as cursor:
    rows = {row[0] for row in cursor}

count = 0
for row in rows:
    count += 1

EndTime = time.clock()
print "Finished in %s seconds" % (EndTime - StartTime)
print "%s features" % count

time.sleep(5) # Let the cpu/ram calm before proceeding!

"""Method 2"""
StartTime2 = time.clock()
arcpy.MakeTableView_management("RandomPoints", "myTableView")
count = int(arcpy.GetCount_management("myTableView").getOutput(0))

EndTime2 = time.clock()
print "Finished in %s seconds" % (EndTime2 - StartTime2)
print "%s features" % count

And some initial results:

>>> 
Finished in 6.75540050237 seconds
1000000 features
Finished in 0.801474780332 seconds
1000000 features
>>> =============================== RESTART ===============================
>>> 
Finished in 6.56968596918 seconds
1000000 features
Finished in 0.812731769756 seconds
1000000 features
>>> =============================== RESTART ===============================
>>> 
Finished in 6.58207512487 seconds
1000000 features
Finished in 0.841122157314 seconds
1000000 features

Imagine larger, more complex datasets. The SearchCursor will indefinitely crawl.

I am not at all dissatisfied with the results, however, the DataAccess module is being used extensively in our GIS development circle. I am looking to rebuild some of our function definitions with this module as it is more flexible than a MakeTableView + GetCount methodology.