RaspberryBASIC.org Forum

[AIR]

Python 3 update.  This is as tight as I can get it at the moment.

Code: Python

1. #!/usr/bin/python3
2.  
3. r = ""
4. a = [None] * 1000001
5. b = bytearray(26)
6.  
7. for x in range(1000001):
8.     b[x % 26] = (x % 26) + 65
9.     a[x] = x
10.     if x % 26 == 25:
11.         r += b.decode()[::-1]
12.  
13. print("r LEN: {}".format(len(r)))
14. print("Front: {}".format(r[:26]))
15. print("Back:  {}".format(r[-26:]))
16. print("UBVal: {}".format(a[1000000]))

riveraa@rpi:~/tmp $ /usr/bin/time ./mil2.py
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
1.32user 0.08system 0:01.41elapsed 99%CPU (0avgtext+0avgdata 27212maxresident)k
0inputs+0outputs (0major+5993minor)pagefaults 0swaps

[AIR]

AIR,

I used the same concept you did for BaCon to speed up its dynamic result 'string'.

Not exactly.  I'm not writing to or reading from a file in the filesystem.  I'm still accessing a string, as an object in memory that exposes methods usually associated with files.

AIR.

[jalih]

The only rule is you can't use fixed size buffers. Storage needs to be dynamic.

Most of the libraries use some sort of pre-defined buffer size and grow that buffer only if capacity is reached. It makes no sense to use fully dynamic buffer as allocating new buffer and copying current buffer contents is a very costly operation to do 38461 times in this challenge!

[John]

Challenge Rules

1. Create an upper case A-Z  character based on the FOR index.

2. Capture the characters in a dynamic reusable string.

3. Maintain a dynamic result string / object / ...

4. Reverse the result.

ScriptBasic is still challenge compliant.

[AIR]

Another Python3 update:

Code: Python

1. #!/usr/bin/python3
2.  
3. def main():
4.     r = ""
5.     a = [None] * 1000001
6.     b = bytearray(26)
7.     decode = bytearray.decode
8.     blah = range(1000001)
9.  
10.     for x in blah:
11.         b[x % 26] = (x % 26) + 65
12.         a[x] = x
13.         if x % 26 == 25:
14.             r += decode(b[::-1])
15.  
16.     print("r LEN: {}".format(len(r)))
17.     print("Front: {}".format(r[:26]))
18.     print("Back:  {}".format(r[-26:]))
19.     print("UBVal: {}".format(a[1000000]))
20.  
21.  
22.  
23. if __name__ == "__main__":
24.     main()

riveraa@rpi:~/tmp $ /usr/bin/time ./mil4.py
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
0.93user 0.13system 0:01.06elapsed 99%CPU (0avgtext+0avgdata 27276maxresident)k
0inputs+0outputs (0major+5997minor)pagefaults 0swaps

AIR.

[AIR]

Last Python update:

Code: Python

1. #!/usr/bin/python3
2.  
3. def main():
4.     t = []
5.     r =""
6.     a = [None] * 1000001
7.     b = bytearray(26)
8.     decode = bytearray.decode
9.     blah = range(1000001)
10.     append = list.append
11.  
12.     for x in blah:
13.         b[x % 26] = (x % 26) + 65
14.         a[x] = x
15.         if x % 26 == 25:
16.             append(t, decode(b[::-1]))
17.  
18.     r = ''.join(t)
19.  
20.     print("r LEN: {}".format(len(r)))
21.     print("Front: {}".format(r[:26]))
22.     print("Back:  {}".format(r[-26:]))
23.     print("UBVal: {}".format(a[1000000]))
24.  
25.  
26.  
27. if __name__ == "__main__":
28.     main()

r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
0.90user 0.09system 0:01.00elapsed 99%CPU (0avgtext+0avgdata 29532maxresident)k
0inputs+0outputs (0major+6508minor)pagefaults 0swaps

AIR.

[John]

Now I'm glad I had that second cup of coffee before updating the Python post. 

Is that doing the reverse on the fly rather than waiting until the result string is built?

You can't get any greyer with that submission.

[AIR]

Yes doing it on the fly, the reversed() function is stupidly inefficient especially with large strings.

I'm doing the reversal with  "[::-1]", after generating the upper case alphabet.

[AIR]

Hope you took the time for a third cup of coffee, John.   

FINAL PYTHON SUBMISSION

Code: Python

1. #!/usr/bin/python3
2.  
3. def main():
4.     t = []
5.     r =""
6.     a = [None] * 1000001
7.     b = bytearray(26)
8.     decode = bytearray.decode
9.     blah = range(1000001)
10.     append = list.append
11.  
12.     for x in blah:
13.         alpha = x % 26
14.         b[alpha] = alpha + 65
15.         a[x] = x
16.         if alpha == 25:
17.             append(t, decode(b[::-1]))
18.  
19.     r = ''.join(t)
20.  
21.     print("r LEN: {}".format(len(r)))
22.     print("Front: {}".format(r[:26]))
23.     print("Back:  {}".format(r[-26:]))
24.     print("UBVal: {}".format(a[1000000]))
25.  
26.  
27.  
28. if __name__ == "__main__":
29.     main()

r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
0.59user 0.09system 0:00.68elapsed 99%CPU (0avgtext+0avgdata 29452maxresident)k
0inputs+0outputs (0major+6390minor)pagefaults 0swaps

I realized that I was calculating the modulo of 'x' multiple times; assigned it to a variable and it made a huge difference!

AIR.

[John]

Saved by a litter box change.

I'll have it posted shortly.


pi@RPi4B:~/python-dev/examples $ timex python3 1mil3-2,py
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
0.68user 0.08system 0:00.78elapsed 99%CPU (0avgtext+0avgdata 29952maxresident)k
0inputs+0outputs (0major+6541minor)pagefaults 0swaps
pi@RPi4B:~/python-dev/examples $

[AIR]

The only rule is you can't use fixed size buffers. Storage needs to be dynamic.

Most of the libraries use some sort of pre-defined buffer size and grow that buffer only if capacity is reached. It makes no sense to use fully dynamic buffer as allocating new buffer and copying current buffer contents is a very costly operation to do 38461 times in this challenge!

You can mitigate some of this if you use some sort of mutable dynamic buffer/array/list/etc.

For my final Python submission, I switched from using straight strings to using a list, which is analogous to an array and added to it in place.  While the size was extended when needed, copying the existing data wasn't needed.

BTW, your 8th submission only runs 1 out of 4 times on my system; the error is the same as before.

Here's the timeing on my RasPi for a successful run:

riveraa@rpi:~/tmp $ timex ./r3
r LEN: 999986
Front: ZYXWVUTSRQPONMLKJIHGFEDCBA
Back:  ZYXWVUTSRQPONMLKJIHGFEDCBA
UBVal: 1000000
1.70user 0.15system 0:01.85elapsed 99%CPU (0avgtext+0avgdata 53088maxresident)k
0inputs+0outputs (0major+12363minor)pagefaults 0swaps

AIR.

[John]

Is a StringBuilder a buffer under the covers that one populates / resets with calls to it?

[John]

I'm waiting for Jalih to show us a 8th submission that lives within the generous rules stated. Exposed buffer code will surely get you busted.

[AIR]

Is a StringBuilder a buffer under the covers that one populates / resets with calls to it?

Basically, yes.

Read this for some insight:  https://www.thecodingdelight.com/java-stringbuilder-class/

[John]

What open the door for me with StringBuilder in the Java submission was the string reverse example I found. Appling it to the rest of the code made all the difference.