C extension with SWIG (counting bytes, not characters)

[rochacbruno]

by @martinxyz

Note: the comparision is not really fair because the C extension is comparing
bytes, while python and rust are comparing utf8 characters.

------------------------------------------------------------------------------------- benchmark: 7 tests ------------------------------------------------------------------------------------
Name (time in us)                 Min                    Max                   Mean              StdDev                 Median                 IQR            Outliers(*)  Rounds  Iterations
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
test_c_bytes_once            962.7170 (1.0)       1,820.2346 (1.0)         989.9444 (1.0)       51.8415 (1.0)         982.5421 (1.0)       11.8364 (1.18)           15;43     721           1
test_rust_once             1,079.2040 (1.12)      1,883.8202 (1.03)      1,101.9418 (1.11)      85.7826 (1.65)      1,087.5929 (1.11)       9.9953 (1.0)            18;67     879           1
test_rust                  2,806.8721 (2.92)      6,881.0866 (3.78)      2,964.6178 (2.99)     360.4035 (6.95)      2,872.8019 (2.92)     183.9128 (18.40)          13;13     339           1
test_regex                28,782.4213 (29.90)    32,900.5020 (18.07)    29,226.6153 (29.52)    705.3509 (13.61)    29,092.1649 (29.61)    194.7829 (19.49)            2;3      34           1
test_pure_python_once     41,820.9783 (43.44)    46,459.0848 (25.52)    42,251.5536 (42.68)    948.0384 (18.29)    42,014.0354 (42.76)    299.1488 (29.93)            1;2      23           1
test_pure_python          56,672.6997 (58.87)    58,843.5191 (32.33)    57,210.8221 (57.79)    588.5420 (11.35)    56,984.3869 (58.00)    250.2394 (25.04)            3;4      18           1
test_itertools            61,560.3020 (63.94)    62,581.4409 (34.38)    61,776.0723 (62.40)    285.4289 (5.51)     61,637.5552 (62.73)    315.8825 (31.60)            2;1      17           1
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

[rochacbruno]

Hey! nice! can you send a Pull Request @martinxyz I would like to add your C implementation to the repo and article

[cuviper]

To do rust_bytes_once, just change its .chars() to .bytes(). I get:

---------------------------------------------------------------------------------------------
Name (time in us)                 Min                    Max                   Mean          
---------------------------------------------------------------------------------------------
test_rust_bytes_once         318.9880 (1.0)         539.6140 (1.0)         319.9789 (1.0)    
test_c_bytes_once            551.8680 (1.73)        891.1050 (1.65)        553.9931 (1.73)   
test_rust_once               694.0350 (2.18)        883.5360 (1.64)        696.2701 (2.18)   
test_rust                  1,806.4350 (5.66)      2,686.6140 (4.98)      1,821.0970 (5.69)   
test_regex                14,234.3260 (44.62)    14,489.9930 (26.85)    14,271.0314 (44.60)  
test_pure_python_once     24,460.7070 (76.68)    32,115.6160 (59.52)    25,180.8352 (78.70)  
test_pure_python          30,797.6000 (96.55)    31,115.7370 (57.66)    30,959.0431 (96.75)  
test_itertools            34,062.9200 (106.78)   35,351.9700 (65.51)    34,241.0481 (107.01) 
---------------------------------------------------------------------------------------------

😄

[rochacbruno]

Thanks @martinxyz and @cuviper

I am going to commit the new implementations and see the results

[martinxyz]

Nice update! By the way, I did not implement bytes because it was faster, but because I had no clue at all how to do this correctly in C. (And I'm not going to try if I can use Rust instead.)