New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Badly selective index could be used for extra filtering even if selective index is used for sorting [CORE5435] #5707
Comments
Modified by: @dyemanovreporter: Dmitry Yemanov [ dimitr ] => Paquito Ines [ paquito ] assignee: Dmitry Yemanov [ dimitr ] Attachment: Databases.zip [ 13052 ] |
Modified by: @dyemanovFix Version: 3.0.2 [ 10785 ] |
Modified by: @dyemanovstatus: Open [ 1 ] => Resolved [ 5 ] resolution: Fixed [ 1 ] Fix Version: 4.0 Alpha 1 [ 10731 ] |
Commented by: Paquito Ines (paquito) We have tested the issue with the Nightly Build, and, though the plan has changed and no longer uses the low selectivity index, the performance has not improved and is even slightly worse. Firebird-3.0.2.32664-0_x64 (nightly build): Maybe is using TABLE1_F1_ID only for ordered retrieval and not for filtering? Forcing the Firebird 2.5 plan works well: PLAN (TABLE1 ORDER TABLE1_F1_ID INDEX (TABLE1_F1_ID)) |
Commented by: Paquito Ines (paquito) Further testing with Firebird 3.0.2.nightly shows a possible problem with the "FIELD1 is null" condition, as it seems to not be using the index for filtering it, but it is working well if the condition is "FIELD1=0", tough both queries have the same plan. Firebird-3.0.2.32664-0_x64 (nightly build): PLAN (TABLE1 ORDER TABLE1_F1_ID) Firebird-3.0.2.32664-0_x64 (nightly build): PLAN (TABLE1 ORDER TABLE1_F1_ID) However, in Firebird 2.5, both "FIELD1 is null" and "FIELD1=0" queries have same plan and same timing: |
Commented by: @dyemanov It appears to be a completely unrelated bug, related to NULLs filtering inside the navigational index scan. I'm investigating. |
Modified by: @pavel-zotovstatus: Resolved [ 5 ] => Resolved [ 5 ] QA Status: No test => Deferred Test Details: Waiting for reply on letter 19.01.2017 08:06. Performance in 3.0.2 and above was improved only for NOT-null values (as it was already noticed here). |
Modified by: @pavel-zotovstatus: Resolved [ 5 ] => Resolved [ 5 ] QA Status: Deferred => Done successfully Test Details: Waiting for reply on letter 19.01.2017 08:06. Performance in 3.0.2 and above was improved only for NOT-null values (as it was already noticed here). => |
Modified by: @pavel-zotovstatus: Resolved [ 5 ] => Closed [ 6 ] |
Commented by: Paquito Ines (paquito) Do you want that we submit a new report for this new bug, so it will be easily tracked? |
Commented by: @dyemanov Please create one. I already have a fix, just need to double check. |
Submitted by: Paquito Ines (paquito)
Attachments:
Databases.zip
It seems that Firebird 3 is sometimes choosing the index with less selectivity, which can have a serious
effect on performance. We have seen this behaviour in several queries.
In this example (we have attached in databases.zip the databases for Firebird 2.5 and Firebird 3):
-Table1 has around 300000 records.
-Field1 has 15000 different values with 20 records each one.
-Field2 has 2 different values (0 with 200000 records, and 1 with 100000 records).
-Both databases were tested immediately after a backup/restore cycle and in the same computer.
-The times are measured in the second execution of each query (though the first execution follow the same pattern).
Query:
select *
from TABLE1 where FIELD1 is null and FIELD2=0
order by FIELD1, ID
Firebird 2.1/2.5:
PLAN (TABLE1 ORDER TABLE1_F1_ID INDEX (TABLE1_F1_ID))
0.002 seconds
Firebird3:
PLAN (TABLE1 ORDER TABLE1_F1_ID INDEX (TABLE1_F2))
Select Expression
-> Filter
-> Table "TABLE1" Access By ID
-> Index "TABLE1_F1_ID" Range Scan (partial match: 1/2)
-> Bitmap
-> Index "TABLE1_F2" Range Scan (full match)
0.240 seconds
-
CREATE TABLE "TABLE1"
(
"ID" INTEGER NOT NULL,
"FIELD1" INTEGER,
"FIELD2" INTEGER,
CONSTRAINT "TABLE1_PK" PRIMARY KEY ("ID")
);
CREATE INDEX "TABLE1_F1_ID" ON "TABLE1"("FIELD1", "ID");
CREATE INDEX "TABLE1_F2" ON "TABLE1"("FIELD2");
Commits: e4f7a6d abc072a
The text was updated successfully, but these errors were encountered: