Percolate index does not search properly by exact phrase query when stemming enabled

[usatenko]

Describe the bug
I use the latest 4.2.0 version.
If I enable stemming mode morphology='stem_en' in combination with e.g. index_exact_words='1' for the percolate index, it does not search properly by exact phrase query.
single parameter morphology='stem_en' searches well, also as index_exact_words='1' without stemming.

To Reproduce
Steps to reproduce the behavior:

Example (not correct behaviour):
1.1

CREATE TABLE jjj (  
title text  
) type='pq' index_exact_words='1' morphology='stem_en'  

2.1

insert into jjj (id, query) values (1, 'finance business partner');  
insert into jjj (id, query) values (2, '="finance business partner"');  
insert into jjj (id, query) values (3, '"finance business partner"');  

3.1

CALL PQ('jjj', '[{"lid": 123, "title": "finance business partner"}]', 'lid' AS docs_id, 1 AS docs);`  
 ------ -----------   
| id   | documents |  
 ------ -----------   
|    1 | 123       |  
|    3 | 123       |  
 ------ -----------   
2 rows in set (0.00 sec)  

But it should give me three IDs: 1, 2 and 3.

Other example (correct behaviour):
1.2

CREATE TABLE eee (  
title text  
) type='pq' morphology='stem_en'  
  

2.2

insert into eee (id, query) values (1, 'finance business partner');  
insert into eee (id, query) values (2, '="finance business partner"');  
insert into eee (id, query) values (3, '"finance business partner"');  
  

3.2

CALL PQ('eee', '[{"lid": 123, "title": "finance business partner"}]', 'lid' AS docs_id, 1 AS docs);  
 ------ -----------   
| id   | documents |  
 ------ -----------   
|    1 | 123       |  
|    2 | 123       |  
|    3 | 123       |  
 ------ -----------   
3 rows in set (0.00 sec)  
  

Other example (correct behaviour):
1.3

CREATE TABLE lll (  
title text  
) type='pq' index_exact_words='1';  

2.3

insert into lll (id, query) values (1, 'finance business partner');  
insert into lll (id, query) values (2, '="finance business partner"');  
insert into lll (id, query) values (3, '"finance business partner"');  
  

3.3

CALL PQ('lll', '[{"lid": 123, "title": "finance business partner"}]', 'lid' AS docs_id, 1 AS docs);  
 ------ -----------   
| id   | documents |  
 ------ -----------   
|    1 | 123       |  
|    2 | 123       |  
|    3 | 123       |  
 ------ -----------   
3 rows in set (0.00 sec)  

Expected behavior
I expect all three percolate index setups to give the same results.

Describe the environment:
Manticore 4.2.0 15e927b28@211223 release (columnar 1.11.4 327b3d4@211223)
Linux manticore-0.local 5.15.0-2-amd64 #1 SMP Debian 5.15.5-2 (2021-12-18) x86_64 GNU/Linux

Messages from log files:
No specific messages related to this

Additional context
N/A

[sanikolaev]

Thank you for the good minimal reproducible example.

[usatenko]

OK, we avoided this bug in our functionality but anyway would be good to have a fix for it

[sanikolaev]

Sure, we'll fix it eventually, it's just that there are more important issues since the functionality which causes this one is not commonly used: percolate + stemming + exact form modifier + phrase operator. So it's going to backlog until we have time to work on it. Pull-requests are very welcome!

[usatenko]

Still exists on Server version: 5.0.2 348514c86@220530 dev git branch HEAD (no branch) Linux manticore-0.local 5.18.0-2-amd64 #1 SMP PREEMPT_DYNAMIC Debian 5.18.5-1 (2022-06-16) x86_64 GNU/Linux

[tomatolog]

I've just fixed percolate query to handle exact term modifier wrong at 3e4d145d

You could install daemon package from the dev repository to get issue fixed after CI finishes packaging