Technical Articles
Eight different sort algorithms implemented in ABAP
- Bucket Sort
- Bubble Sort
- Merge Sort
- Quick Sort
- Selection Sort
- Insertion Sort
- Heap Sort
- Shell Sort
- A very draft performance comparison
- Sleep Sort in JavaScript
Some application developers think that it is enough to know SORT keyword and how to use sorted table in ABAP for their daily work without knowing how SORT is done internally. For me I can not say this assumption is wrong. I personal preference is to know something more thoroughly. We have learned various sort algorithms in the university, here I just list my implementation on some of them using ABAP for my personal study purpose.
For each sort algorithm I will create a static public class with a sort method which accepts an internal table with unsorted Integer and an output table which are sorted. For simplification reason the element in the internal table only consists of unsigned integers ( >= 0 )
Bucket Sort ( In China we prefer to call it Hash Sort )
In ABAP the internal table is a perfect choice for bucket collection 🙂 A small trap here is, array in most program language has start index as 0, however in ABAP for internal table it is 1. So be careful about the possibility that 0 appears in the input internal table.
And I also implement a version using JavaScript which can support negative integer in Bucket Sort as well. See source code here.
Bubble Sort
I have implemented two variants, the only difference between them:
- Variant 1 uses two nested DO LOOP, while variant 2 uses WHILE as inner LOOP.
- Variant 1 uses traditional keyword MODIFY itab FROM workarea INDEX index to swap the two adjacent element, while variant 2 uses new grammar itab[ index ].
Source code for both variants.
Merge Sort
Again I have implemented two variants.
Variant1 – use Recursive
Callstack could be found below:
Variant 2 – non recursive version
This variant is implemented in a non-recursive way.
Source code of both variants.
Quick Sort
Selection Sort
Insertion Sort
Heap Sort
Shell Sort
A very draft performance comparison
Since each sort algorithm has different time complexity – best case, worst case and average case according to different data distribution, here below I only make a very draft comparison by generating some random integers in ABAP via cl_abap_random_int:
DATA: lv_seed TYPE i.
lv_seed = sy-timlo.
DATA(lo_ran) = cl_abap_random_int=>create( min = 1 max = 1000 seed = lv_seed ).
DO iv_num TIMES.
APPEND lo_ran->get_next( ) TO rv_table.
ENDDO.Meanwhile I am especially curious about how ABAP keyword SORT will behave against these eight sort algorithms, so I create another two sort approaches.
The ninth sort approach
Pretty simple, just use ABAP keyword SORT to sort the table.
method SORT.
rv_table = iv_table.
SORT rv_table.
endmethod.​The tenth sort approach
I just loop the original table and put each element to a sorted table with line item as INT4.
DATA: lt_sorted TYPE ZTSORTED_INT4.
LOOP AT iv_table ASSIGNING FIELD-SYMBOL(<item>).
INSERT <item> INTO table lt_sorted.
ENDLOOP.
APPEND LINES OF lt_sorted TO rv_table.The table type ZTSORTED_INT4 has line type INT4 with sorted table type.
Test code:
DATA(lt_test_data) = zcl_sort_helper=>generate_data( 3000 ).
zcl_sort_helper=>start_measure( ).
DATA(lt_bubble) = zcl_bubblesort=>sort( lt_test_data ).
WRITE: / 'Bubble Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_hashsort) = zcl_hashsort=>sort( lt_test_data ).
WRITE: / 'Hash Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_heapsort) = zcl_hashsort=>sort( lt_test_data ).
WRITE: / 'Heap Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_insertsort) = zcl_insertsort=>sort( lt_test_data ).
WRITE: / 'Insert Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_mergesort) = zcl_mergesort=>sort( lt_test_data ).
WRITE: / 'Merge Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_quicksort) = zcl_quicksort=>sort( lt_test_data ).
WRITE: / 'Quick Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_selectsort) = zcl_selectsort=>sort( lt_test_data ).
WRITE: / 'Select Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_shellsort) = zcl_shellsort=>sort( lt_test_data ).
WRITE: / 'Shell Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_sort_keyword) = zcl_sort_via_keyword=>sort( lt_test_data ).
WRITE: / 'ABAP Sort keyword duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_sort_table) = zcl_abap_sorttable=>sort( lt_test_data ).
WRITE: / 'ABAP Sorted table duration:' , zcl_sort_helper=>stop( ).
ASSERT lt_bubble = lt_hashsort.
ASSERT lt_hashsort = lt_heapsort.
ASSERT lt_heapsort = lt_insertsort.
ASSERT lt_insertsort = lt_mergesort.
ASSERT lt_mergesort = lt_quicksort.
ASSERT lt_quicksort = lt_selectsort.
ASSERT lt_shellsort = lt_selectsort.
ASSERT lt_sort_keyword = lt_shellsort.
ASSERT lt_sort_table = lt_sort_keyword.The test result ( unit: microsecond )
The ABAP SORT keyword and SORTED TABLE did a really good job here 🙂
The complete source code for this blog could be found from my github.
Sleep Sort in JavaScript
Last but not least, the super cool “Sleep Sort” done in JavaScript, which does not need any comparison against two elements in the array.
const num = [1,5,6,11,2,3,4,8,7,14];
num.forEach( num => {
setTimeout( () => { console.log(num)}, num);
});Test output:
Happy coding 🙂
update on 2022-3-26
the latest url is here: https://github.com/wangzixi-diablo/ui5-tutorial/tree/main/abap/sort
Thanks for sharing with us , it is really helpful.
Hi, Could you share the links/attachments if you have it. Thank you....
sorry, the new url is here: https://github.com/wangzixi-diablo/ui5-tutorial/tree/main/abap/sort
Thanks for sharing.
However, I must wonder: Are there any real use cases for manual sort in ABAP?
Hi Shai,
As far as I know, the answer is no. As least for me, I never use the manual sort in my application development, the keyword SORT does a good job enough. I implement these eight algorithm for fun, just in order to refresh what I have learned in my university. In China when graduates are looking for job, they are tended to be asked to write implementation out in paper or white board 🙂
Best regards,
Jerry
Hi Jerry, Could you share the links/attachments which would really helpful. Thank you…
sorry, the new url is here: https://github.com/wangzixi-diablo/ui5-tutorial/tree/main/abap/sort
Thanks for sharing. Good thought...
Hi Joy, Could you share the links/attachments if you have it. Really appreciate your help...Thank you…
Thanks for sharing! Very cool!!! .....and it really bothers me when people as "yeh, but why are you doing this? when would you ever use this is in the real world?"......Hey Debbie Downer, what is wrong with doing it for FUN ??!?! I know....crazy concept, right?!?! Some people actually do have fun just coding and seeing what they can create/re-create/do with it.......so kudos to you for that!!!
thanks for the explanation
Thank you very much for sharing such wonderful information. Especially comparison at the end is awesome. Actually i am wondering what algorithm Sort Keyword will be using to sort data. Do you have any clue about it.
Hi Dilip, Could you share the links/attachments if you have it. Thank you….
sorry, the new url is here: https://github.com/wangzixi-diablo/ui5-tutorial/tree/main/abap/sort
Hi, The git links here are not working.
Hi Jerry,
Nice blog. Thanks for sharing. It seems github links is not work, do have any new links for your code example?
Thanks,
Harish
Hello Git links are not working? new links please
Hello Jerry,
Git Link is not working. Can you please share the working link.
Git Link is not working. Can you please share the working link.
Hi dear,
sorry for some reason, my original github repo got deleted. Please refer to this new repo for source code:Â https://github.com/wangzixi-diablo/abap/tree/master/ABAP
Best regards,
Jerry
We lost this repo now. Could you please share the updated one please.
Hi Jerry, can you share the working Github link, please.
Hi dear,
sorry for some reason, my original github repo got deleted. Please refer to this new repo for source code:Â https://github.com/wangzixi-diablo/abap/tree/master/ABAP
Best regards,
Jerry
Hi Jerry,
Would appreciate if you can please provide updated links for each implementation.
Thanks
Pradipta
Hi dear,
sorry for some reason, my original github repo got deleted. Please refer to this new repo for source code: https://github.com/wangzixi-diablo/abap/tree/master/ABAP
Best regards,
Jerry
Thank you for sharing the new link, Jerry !
Thank you so much for detailed information. Could you share the GitHub links as the existing ones are not working. Thank you...
Hi dear,
sorry for some reason, my original github repo got deleted. Please refer to this new repo for source code: https://github.com/wangzixi-diablo/abap/tree/master/ABAP
Best regards,
Jerry
Hi Jerry,
We lost the repo again. Could you please check and share the new one?
Karthik
sorry, the new url is here: https://github.com/wangzixi-diablo/ui5-tutorial/tree/main/abap/sort
Thanks a lot Jerry..
Nice work! Thank you very much.
In your ZSORT_TEST.abap, I find codes like these:
zcl_sort_helper=>start_measure( ).
DATA(lt_hashsort) = zcl_hashsort=>sort( lt_test_data ).
WRITE: / 'Hash Sort duration:' , zcl_sort_helper=>stop( ).
zcl_sort_helper=>start_measure( ).
DATA(lt_heapsort) = zcl_hashsort=>sort( lt_test_data ).
WRITE: / 'Heap Sort duration:' , zcl_sort_helper=>stop( ).
I think in the second group, " zcl_hashsort " should be " zcl_heapsort ".
I guess, that it is the reason why hashsort is almost as fast as heapsort in your result.
Anyway, I learn a lot from your work.
Many thanks.
Good to see the algorithm !
Regarding the Quick Sort algo, its based on divide & conquer algorithm using a pivot element as per the link provided by you and Wiki.
I do not see a pivot selection nor divide & conquer logic in the code. But I see swapping logic which is part of Bubble Sort & Insertion Sort algorithms.