Currently the code looks something like this:
LOOP AT lt_orders ASSIGNING <fs_order>.
SELECT COUNT(*) AS cnt
FROM order_items
INTO <fs_order>-cnt
WHERE order_id = <fs_order>-order_id.
ENDLOOP.
It is the slowest part of the report. I want to speed it up.
How can I use FOR ALL ENTRIES with GROUP BY?
Check the documentation. You can't use GROUP BY. Maybe in this case, you could try selecting your items with FAE outside of the loop, then count them using a parallel cursor:
REPORT.
TYPES: BEGIN OF ty_result,
vbeln TYPE vbeln,
cnt TYPE i.
TYPES: END OF ty_result.
DATA: lt_headers TYPE SORTED TABLE OF ty_result WITH UNIQUE KEY vbeln,
lv_tabix TYPE sy-tabix VALUE 1.
"get the headers
SELECT vbeln FROM vbak UP TO 100 ROWS INTO CORRESPONDING FIELDS OF TABLE lt_headers.
"get corresponding items
SELECT vbeln, posnr FROM vbap FOR ALL ENTRIES IN #lt_headers
WHERE vbeln EQ #lt_headers-vbeln
ORDER BY vbeln, posnr
INTO TABLE #DATA(lt_items).
LOOP AT lt_headers ASSIGNING FIELD-SYMBOL(<h>).
LOOP AT lt_items FROM lv_tabix ASSIGNING FIELD-SYMBOL(<i>).
IF <i>-vbeln NE <h>-vbeln.
lv_tabix = sy-tabix.
EXIT.
ELSE.
<h>-cnt = <h>-cnt + 1.
ENDIF.
ENDLOOP.
ENDLOOP.
BREAK-POINT.
Or join header/item with a distinct count on the item id (whichever column that would be in your table).
You should be able to do something like
SELECT COUNT(order_item_id) AS cnt, order_id
FROM order_items
INTO CORRESPONDING FIELDS OF TABLE lt_count
GROUP BY order_id.
Assuming that order_item_id is a key in the order_items table. And assuming that lt_count has two fields: cnt of type int8 and order_id of same type as your other order_id fields
PS: then you can loop over lt_count and move the counts to lt_orders. Or the other way around. To speed up the loop, sort one of the tables and use READ ... BINARY SEARCH
I did with table KNB1 (customer master in company code), where we have customers, which are created in several company codes.
Please note, because of FOR ALL ENTRIES you have to SELECT the full key.
TYPES: BEGIN OF ty_knb1,
kunnr TYPE knb1-kunnr,
count TYPE i,
END OF ty_knb1.
TYPES: BEGIN OF ty_knb1_fae,
kunnr TYPE knb1-kunnr,
END OF ty_knb1_fae.
DATA: lt_knb1_fae TYPE STANDARD TABLE OF ty_knb1_fae.
DATA: lt_knb1 TYPE HASHED TABLE OF ty_knb1
WITH UNIQUE KEY kunnr.
DATA: ls_knb1 TYPE ty_knb1.
DATA: ls_knb1_db TYPE knb1.
START-OF-SELECTION.
lt_knb1_fae = VALUE #( ( kunnr = ... ) ). "add at least one customer which is created in several company codes
ls_knb1-count = 1.
SELECT kunnr bukrs
INTO CORRESPONDING FIELDS OF ls_knb1_db
FROM knb1
FOR ALL ENTRIES IN lt_knb1_fae
WHERE kunnr EQ lt_knb1_fae-kunnr.
ls_knb1-kunnr = ls_knb1_db-kunnr.
COLLECT ls_knb1 INTO lt_knb1.
ENDSELECT.
Create a range table for your lt_orders, like lt_orders_range.
Do select order_id, count( * ) where order_id in lt_orders_range.
If you think this is too much to create a range table, you will save a lot of performance by running just one select for all orders instead of single select for each order id.
Not directly, only through a CDS view
While all of the answers provide a faster solution than the one in the question, the fastest way is not mentioned.
If you have at least Netweaver 7.4, EHP 5 (and you should, it was released in 2014), you can use CDS views, even if you are not on HANA.
It still cannot be done directly, as OpenSQL does not allow FOR ALL ENTRIES with GROUP BY, and CDS views cannot handle FOR ALL ENTRIES. However, you can create one of each.
CDS:
#AbapCatalog.sqlViewName: 'zorder_i_fae'
DEFINE VIEW zorder_items_fae AS SELECT FROM order_items {
order_id,
count( * ) AS cnt,
}
GROUP BY order_id
OpenSQL:
SELECT *
FROM zorder_items_fae
INTO TABLE #DATA(lt_order_cnt)
FOR ALL ENTRIES IN #lt_orders
WHERE order_id = #lt_orders-order_id.
Speed
If lt_orders contains more than about 30% of all possible order_id values from table ORDER_ITEMS, the answer from iPirat is faster. (While using more memory, obviously)
However, if you need only a couple hunderd order_id values out of millions, this solution is about 10 times faster than any other answer, and 100 times faster than the original.
Related
I have a table with >1M rows of data and 20+ columns.
Within my table (tableX) I have identified duplicate records (~80k) in one particular column (troubleColumn).
If possible I would like to retain the original table name and remove the duplicate records from my problematic column otherwise I could create a new table (tableXfinal) with the same schema but without the duplicates.
I am not proficient in SQL or any other programming language so please excuse my ignorance.
delete from Accidents.CleanedFilledCombined
where Fixed_Accident_Index
in(select Fixed_Accident_Index from Accidents.CleanedFilledCombined
group by Fixed_Accident_Index
having count(Fixed_Accident_Index) >1);
You can remove duplicates by running a query that rewrites your table (you can use the same table as the destination, or you can create a new table, verify that it has what you want, and then copy it over the old table).
A query that should work is here:
SELECT *
FROM (
SELECT
*,
ROW_NUMBER()
OVER (PARTITION BY Fixed_Accident_Index)
row_number
FROM Accidents.CleanedFilledCombined
)
WHERE row_number = 1
UPDATE 2019: To de-duplicate rows on a single partition with a MERGE, see:
https://stackoverflow.com/a/57900778/132438
An alternative to Jordan's answer - this one scales better when having too many duplicates:
#standardSQL
SELECT event.* FROM (
SELECT ARRAY_AGG(
t ORDER BY t.created_at DESC LIMIT 1
)[OFFSET(0)] event
FROM `githubarchive.month.201706` t
# GROUP BY the id you are de-duplicating by
GROUP BY actor.id
)
Or a shorter version (takes any row, instead of the newest one):
SELECT k.*
FROM (
SELECT ARRAY_AGG(x LIMIT 1)[OFFSET(0)] k
FROM `fh-bigquery.reddit_comments.2017_01` x
GROUP BY id
)
To de-duplicate rows on an existing table:
CREATE OR REPLACE TABLE `deleting.deduplicating_table`
AS
# SELECT id FROM UNNEST([1,1,1,2,2]) id
SELECT k.*
FROM (
SELECT ARRAY_AGG(row LIMIT 1)[OFFSET(0)] k
FROM `deleting.deduplicating_table` row
GROUP BY id
)
Not sure why nobody mentioned DISTINCT query.
Here is the way to clean duplicate rows:
CREATE OR REPLACE TABLE project.dataset.table
AS
SELECT DISTINCT * FROM project.dataset.table
If your schema doesn’t have any records - below variation of Jordan’s answer will work well enough with writing over same table or new one, etc.
SELECT <list of original fields>
FROM (
SELECT *, ROW_NUMBER() OVER (PARTITION BY Fixed_Accident_Index) AS pos,
FROM Accidents.CleanedFilledCombined
)
WHERE pos = 1
In more generic case - with complex schema with records/netsed fields, etc. - above approach can be a challenge.
I would propose to try using Tabledata: insertAll API with rows[].insertId set to respective Fixed_Accident_Index for each row.
In this case duplicate rows will be eliminated by BigQuery
Of course, this will involve some client side coding - so might be not relevant for this particular question.
I havent tried this approach by myself either but feel it might be interesting to try :o)
If you have a large-size partitioned table, and only have duplicates in a certain partition range. You don't want to overscan nor process the whole table. use the MERGE SQL below with predicates on partition range:
-- WARNING: back up the table before this operation
-- FOR large size timestamp partitioned table
-- -------------------------------------------
-- -- To de-duplicate rows of a given range of a partition table, using surrage_key as unique id
-- -------------------------------------------
DECLARE dt_start DEFAULT TIMESTAMP("2019-09-17T00:00:00", "America/Los_Angeles") ;
DECLARE dt_end DEFAULT TIMESTAMP("2019-09-22T00:00:00", "America/Los_Angeles");
MERGE INTO `gcp_project`.`data_set`.`the_table` AS INTERNAL_DEST
USING (
SELECT k.*
FROM (
SELECT ARRAY_AGG(original_data LIMIT 1)[OFFSET(0)] k
FROM `gcp_project`.`data_set`.`the_table` AS original_data
WHERE stamp BETWEEN dt_start AND dt_end
GROUP BY surrogate_key
)
) AS INTERNAL_SOURCE
ON FALSE
WHEN NOT MATCHED BY SOURCE
AND INTERNAL_DEST.stamp BETWEEN dt_start AND dt_end -- remove all data in partiion range
THEN DELETE
WHEN NOT MATCHED THEN INSERT ROW
credit: https://gist.github.com/hui-zheng/f7e972bcbe9cde0c6cb6318f7270b67a
Easier answer, without a subselect
SELECT
*,
ROW_NUMBER()
OVER (PARTITION BY Fixed_Accident_Index)
row_number
FROM Accidents.CleanedFilledCombined
WHERE TRUE
QUALIFY row_number = 1
The Where True is neccesary because qualify needs a where, group by or having clause
Felipe's answer is the best approach for most cases. Here is a more elegant way to accomplish the same:
CREATE OR REPLACE TABLE Accidents.CleanedFilledCombined
AS
SELECT
Fixed_Accident_Index,
ARRAY_AGG(x LIMIT 1)[SAFE_OFFSET(0)].* EXCEPT(Fixed_Accident_Index)
FROM Accidents.CleanedFilledCombined AS x
GROUP BY Fixed_Accident_Index;
To be safe, make sure you backup the original table before you run this ^^
I don't recommend to use ROW NUMBER() OVER() approach if possible since you may run into BigQuery memory limits and get unexpected errors.
Update BigQuery schema with new table column as bq_uuid making it NULLABLE and type STRING
Create duplicate rows by running same command 5 times for example
insert into beginner-290513.917834811114.messages (id, type, flow, updated_at) Values(19999,"hello", "inbound", '2021-06-08T12:09:03.693646')
Check if duplicate entries exist
select * from beginner-290513.917834811114.messages where id = 19999
Use generate uuid function to generate uuid corresponding to each message
UPDATE beginner-290513.917834811114.messages
SET bq_uuid = GENERATE_UUID()
where id>0
Clean duplicate entries
DELETE FROM beginner-290513.917834811114.messages
WHERE bq_uuid IN
(SELECT bq_uuid
FROM
(SELECT bq_uuid,
ROW_NUMBER() OVER( PARTITION BY updated_at
ORDER BY bq_uuid ) AS row_num
FROM beginner-290513.917834811114.messages ) t
WHERE t.row_num > 1 );
I have a table with 3 columns:
table1: ID, CODE, RESULT, RESULT2, RESULT3
I have this SAS code:
data table1
set table1;
BY ID, CODE;
IF FIRST.CODE and RESULT='A' THEN OUTPUT;
ELSE IF LAST.CODE and RESULT NE 'A' THEN OUTPUT;
RUN;
So we are grouping the data by ID and CODE, and then writing to the dataset if certain conditions are met. I want to write a hive query to replicate this. This is what I have:
proc sql;
create table temp as
select *, row_number() over (partition by ID, CODE) as rowNum
from table1;
create table temp2 as
select a.ID, a.CODE, a.RESULT, a.RESULT2, a.RESULT3
from temp a
inner join (select ID, CODE, max(rowNum) as maxRowNum
from temp
group by ID, CODE) b
on a.ID=b.ID and a.CODE=b.CODE
where (a.rowNum=1 and a.RESULT='A') or (a.rowNum=b.maxRowNum and a.RESULT NE 'A');
quit;
There are two issues I see with this.
1) The row that is first or last in each BY group is entirely dependant on the order of rows in table1 in SAS, we aren't ordering by anything. I don't think row order is preserved when translating to a hive query.
2) The SAS code is taking the first row in each BY GROUP or the last, not both. I think that my HIVE query is taking both, resulting in more rows than I want.
Any suggestions or insight on how to improve my query is appreciated. Is it even possible to replicate this SAS code in HIVE?
The SAS code has a by statement (BY ID CODE;), which tells SAS that the set dataset is sorted at those levels. So, not a random selection for first. and last..
That said, we can replicate this in HIVE by using the first_value and last_value window functions.
FIRST.CODE should replicate to
first_value(code) over (partition by Id order by code)fcode
Similarly, LAST.CODE would be
last_value(code) over (partition by Id order by code)lcode
Once you have the fcode and lcode columns, use case when statements for the result column criteria. Like,
case when (code=fcode and result='A') or (code=lcode and result<>'A')
then 1 else 0 end as op_flag
Then the fetch the table with where op_flag = 1
SAMPLE
select id, code, result from (
select *,
first_value(code) over (partition by id order by code)fcode,
last_value(code) over (partition by id order by code)lcode
from footab) f
where (code=fcode and result='A') or (code=lcode and result<>'A')
Regarding point 1) the BY group processing requires the input data to be sorted or indexed on BY variables, so though the code contains no ordering, the source data is processed in order. If the input data was not indexed/sorted, SAS will throw error.
Regarding this, possible differences are on rows with same values of BY variables, especially if the RESULT is different.
In SAS, I would pre-sort data by ID, CODE, RESULT, then use BY ID CODE in order to not be influenced by order of rows.
Regarding 2) FIRST and LAST can be both true in SAS. Since your condition for first and last on RESULT is different, I guess this is not a source of differences.
I guess you could add another field as
row_number() over (partition by ID, CODE desc) as rowNumDesc
to detect last row with rowNumDesc = 1 (so that you skip the join).
EDIT:
I think the two programs above both include random selection of rows for groups with same values of ID and CODE variables, especially with same values of RESULT. But you should get same number of rows from both. If not, just debug it.
However the random aspect in SAS code/storage is based on physical order of rows, while the ROW_NUMBERs randomness within a group will be influenced by the implementation of the function in the engine.
The easiest way to ask my question is with a Hypothetical Scenario.
Lets say we have 3 tables. Singapore_Prices, Produce_val, and Bosses_unreasonable_demands.
So Prices is a pretty simple table. Item column containing a name, and a Price column containing a number.
Produce_Val is also simple 2 column table. Type column containing what type the produce is (Fruit or veggie) and then Name column (Tomato, pineapple, etc.)
The Bosses_unreasonable_demands only contains one column, Fruit, which CAN contain the names of some fruits.
OK? Ok.
SO, My boss wants me to write a query that returns the prices for every fruit in his unreasonable demands table. Simple enough. BUT, if he doesn't have any entries in his table, he just wants me to output the prices of ALL fruits that exist in produce_val.
Now, assuming I don't know where the DBA who designed this silly hypothetical system lives (and therefore can't get him to fix this), our query would look like this:
if <Logic to determine if Bosses demands are empty>
Then
select Item, Price
from Singapore_Prices
where Item in (select Fruit from Bosses_Unreasonable_demands)
Else
select Item, Price
from Singapore_Prices
where Item in (select Name from Produce_val where type = 'Fruit')
end if;
(Well, we'd select those into a variable, and then output the variable, probably with bulk-collect shenanigans, but that's not important)
Which works. It is entirely functional, and won't be slow, even if we extend it out to 2000 other stores other than Singapore. (Well, no slower than anything else that touches 2000 some tables) BUT, I'm still doing two different select statements that are practically identical. My Comp Sci teacher rolls in their grave every time my fingers hit ctrl-V. I can cut this code in half and only do one select statement. I KNOW I can.
I just have no earthly idea how. I can't use cursors as an in statement, I can't use nested tables or varrays, I can't use cleverly crafted strings, I... I just... I don't know. I don't know how to do this. Is there a way? Does it exist?
Or do I have to copy/paste forever?
Your best bet would be dynamic SQL, because you can't parameterize table or column names.
You will have a SQL query template, have a logic to determine tables and columns that you want to query, then blend them together and execute.
Another aproach, (still a lot of ctrl-v like code) is to use set construction UNION ALL:
select 1st query where boss_condition
union all
select 2nd query where not boss_condition
Try this:
SELECT *
FROM (SELECT s.*, 'BOSS' AS FRUIT_SOURCE
FROM BOSSES_UNREASONABLE_DEMANDS b
INNER JOIN SINGAPORE_FRUIT_LIST s
ON s.ITEM = b.FRUIT
CROSS JOIN (SELECT COUNT(*) AS BOSS_COUNT
FROM BOSSES_UNREASONABLE_DEMANDS)) x
UNION ALL
(SELECT s.*, 'NORMAL' AS FRUIT_SOURCE
FROM PRODUCE_VAL p
INNER JOIN SINGAPORE_FRUIT_LIST s
ON (s.ITEM = p.NAME AND
s.TYPE = 'Fruit')
CROSS JOIN (SELECT COUNT(*) AS BOSS_COUNT
FROM BOSSES_UNREASONABLE_DEMANDS)) n
WHERE (BOSS_COUNT > 0 AND FRUIT_SOURCE = 'BOSS') OR
(BOSS_COUNT = 0 AND FRUIT_SOURCE = 'NORMAL')
Share and enjoy.
I think you can use nested tables. Assume you have a schema-level nested table type FRUIT_NAME_LIST (defined using CREATE TYPE).
SELECT fruit
BULK COLLECT INTO my_fruit_name_list
FROM bosses_unreasonable_demands
;
IF my_fruit_name_list.count = 0 THEN
SELECT name
BULK COLLECT INTO my_fruit_name_list
FROM produce_val
WHERE type='Fruit'
;
END IF;
SELECT item, price
FROM singapore_prices
WHERE item MEMBER OF my_fruit_name_list
;
(or, WHERE item IN (SELECT column_value FROM TABLE(CAST(my_fruit_name_list AS fruit_name_list)) if you like that better)
I think I could use some help here from more experienced users...
I have an integer field name in a table, let's call it SO_ID in a table SO, and to each new row I need to calculate a new SO_ID based on the following rules
1) SO_ID consists of 6 letters where first 3 are an area code, and the last three is the sequenced number within this area.
309001
309002
309003
2) so the next new row will have a SO_ID of value
309004
3) if someone deletes the row with SO_ID value = 309002, then the next new row must recycle this value, so the next new row has got to have the SO_ID of value
309002
can anyone please provide me with either a SQL function or PL/SQL (perhaps a trigger straightaway?) function that would return the next available SO_ID I need to use ?
I reckon I could get use of keyword rownum in my sql, but the follwoing just doens't work properly
select max(so_id),max(rownum) from(
select (so_id),rownum,cast(substr(cast(so_id as varchar(6)),4,3) as int) from SO
where length(so_id)=6
and substr(cast(so_id as varchar(6)),1,3)='309'
and cast(substr(cast(so_id as varchar(6)),4,3) as int)=rownum
order by so_id
);
thank you for all your help!
This kind of logic is fraught with peril. What if two sessions calculate the same "next" value, or both try to reuse the same "deleted" value? Since your column is an integer, you'd probably be better off querying "between 309001 and 309999", but that begs the question of what happens when you hit the thousandth item in area 309?
Is it possible to make SO_ID a foreign key to another table as well as a unique key? You could pre-populate the parent table with all valid IDs (or use a function to generate them as needed), and then it would be a simple matter to select the lowest one where a child record doesn't exist.
well, we came up with this... sort of works.. concurrency is 'solved' via unique constraint
select min(lastnumber)
from
(
select so_id,so_id-LAG(so_id, 1, so_id) OVER (ORDER BY so_id) AS diff,LAG(so_id, 1, so_id) OVER (ORDER BY so_id)as lastnumber
from so_miso
where substr(cast(so_id as varchar(6)),1,3)='309'
and length(so_id)=6
order by so_id
)a
where diff>1;
Do you really need to compute & store this value at the time a row is inserted? You would normally be better off storing the area code and a date in a table and computing the SO_ID in a view, i.e.
SELECT area_code ||
LPAD( DENSE_RANK() OVER( PARTITION BY area_code
ORDER BY date_column ),
3,
'0' ) AS so_id,
<<other columns>>
FROM your_table
or having a process that runs periodically (nightly, for example) to assign the SO_ID using similar logic.
If your application is not pure sql, you could do this in application code (ie: Java code). This would be more straightforward.
If you are recycling numbers when rows are deleted, your base table must be consulted when generating the next number. "Legacy" pre-relational schemes that attempt to encode information in numbers are a pain to make airtight when numbers must be recycled after deletes, as you say yours must.
If you want to avoid having to scan your table looking for gaps, an after-delete routine must write the deleted number to a separate table in a "ReuseMe" column. The insert routine does this:
begins trans
selects next-number table for update
uses a reuseme number if available else uses the next number
clears the reuseme number if applicable or increments the next-number in the next-number table
commits trans
Ignoring the issues about concurrency, the following should give a decent start.
If 'traffic' on the table is low enough, go with locking the table in exclusive mode for the duration of the transaction.
create table blah (soc_id number(6));
insert into blah select 309000 + rownum from user_tables;
delete from blah where soc_id = 309003;
commit;
create or replace function get_next (i_soc in number) return number is
v_min number := i_soc* 1000;
v_max number := v_min + 999;
begin
lock table blah in exclusive mode;
select min(rn) into v_min
from
(select rownum rn from dual connect by level <= 999
minus
select to_number(substr(soc_id,4))
from blah
where soc_id between v_min and v_max);
return v_min;
end;
Have the following tables (Oracle 10g):
catalog (
id NUMBER PRIMARY KEY,
name VARCHAR2(255),
owner NUMBER,
root NUMBER REFERENCES catalog(id)
...
)
university (
id NUMBER PRIMARY KEY,
...
)
securitygroup (
id NUMBER PRIMARY KEY
...
)
catalog_securitygroup (
catalog REFERENCES catalog(id),
securitygroup REFERENCES securitygroup(id)
)
catalog_university (
catalog REFERENCES catalog(id),
university REFERENCES university(id)
)
Catalog: 500 000 rows, catalog_university: 500 000, catalog_securitygroup: 1 500 000.
I need to select any 50 rows from catalog with specified root ordered by name for current university and current securitygroup. There is a query:
SELECT ccc.* FROM (
SELECT cc.*, ROWNUM AS n FROM (
SELECT c.id, c.name, c.owner
FROM catalog c, catalog_securitygroup cs, catalog_university cu
WHERE c.root = 100
AND cs.catalog = c.id
AND cs.securitygroup = 200
AND cu.catalog = c.id
AND cu.university = 300
ORDER BY name
) cc
) ccc WHERE ccc.n > 0 AND ccc.n <= 50;
Where 100 - some catalog, 200 - some securitygroup, 300 - some university. This query return 50 rows from ~ 170 000 in 3 minutes.
But next query return this rows in 2 sec:
SELECT ccc.* FROM (
SELECT cc.*, ROWNUM AS n FROM (
SELECT c.id, c.name, c.owner
FROM catalog c
WHERE c.root = 100
ORDER BY name
) cc
) ccc WHERE ccc.n > 0 AND ccc.n <= 50;
I build next indexes: (catalog.id, catalog.name, catalog.owner), (catalog_securitygroup.catalog, catalog_securitygroup.index), (catalog_university.catalog, catalog_university.university).
Plan for first query (using PLSQL Developer):
http://habreffect.ru/66c/f25faa5f8/plan2.jpg
Plan for second query:
http://habreffect.ru/f91/86e780cc7/plan1.jpg
What are the ways to optimize the query I have?
The indexes that can be useful and should be considered deal with
WHERE c.root = 100
AND cs.catalog = c.id
AND cs.securitygroup = 200
AND cu.catalog = c.id
AND cu.university = 300
So the following fields can be interesting for indexes
c: id, root
cs: catalog, securitygroup
cu: catalog, university
So, try creating
(catalog_securitygroup.catalog, catalog_securitygroup.securitygroup)
and
(catalog_university.catalog, catalog_university.university)
EDIT:
I missed the ORDER BY - these fields should also be considered, so
(catalog.name, catalog.id)
might be beneficial (or some other composite index that could be used for sorting and the conditions - possibly (catalog.root, catalog.name, catalog.id))
EDIT2
Although another question is accepted I'll provide some more food for thought.
I have created some test data and run some benchmarks.
The test cases are minimal in terms of record width (in catalog_securitygroup and catalog_university the primary keys are (catalog, securitygroup) and (catalog, university)). Here is the number of records per table:
test=# SELECT (SELECT COUNT(*) FROM catalog), (SELECT COUNT(*) FROM catalog_securitygroup), (SELECT COUNT(*) FROM catalog_university);
?column? | ?column? | ?column?
----------+----------+----------
500000 | 1497501 | 500000
(1 row)
Database is postgres 8.4, default ubuntu install, hardware i5, 4GRAM
First I rewrote the query to
SELECT c.id, c.name, c.owner
FROM catalog c, catalog_securitygroup cs, catalog_university cu
WHERE c.root < 50
AND cs.catalog = c.id
AND cu.catalog = c.id
AND cs.securitygroup < 200
AND cu.university < 200
ORDER BY c.name
LIMIT 50 OFFSET 100
note: the conditions are turned into less then to maintain comparable number of intermediate rows (the above query would return 198,801 rows without the LIMIT clause)
If run as above, without any extra indexes (save for PKs and foreign keys) it runs in 556 ms on a cold database (this is actually indication that I oversimplified the sample data somehow - I would be happier if I had 2-4s here without resorting to less then operators)
This bring me to my point - any straight query that only joins and filters (certain number of tables) and returns only a certain number of the records should run under 1s on any decent database without need to use cursors or to denormalize data (one of these days I'll have to write a post on that).
Furthermore, if a query is returning only 50 rows and does simple equality joins and restrictive equality conditions it should run even much faster.
Now let's see if I add some indexes, the biggest potential in queries like this is usually the sort order, so let me try that:
CREATE INDEX test1 ON catalog (name, id);
This makes execution time on the query - 22ms on a cold database.
And that's the point - if you are trying to get only a page of data, you should only get a page of data and execution times of queries such as this on normalized data with proper indexes should take less then 100ms on decent hardware.
I hope I didn't oversimplify the case to the point of no comparison (as I stated before some simplification is present as I don't know the cardinality of relationships between catalog and the many-to-many tables).
So, the conclusion is
if I were you I would not stop tweaking indexes (and the SQL) until I get the performance of the query to go below 200ms as rule of the thumb.
only if I would find an objective explanation why it can't go below such value I would resort to denormalisation and/or cursors, etc...
First I assume that your University and SecurityGroup tables are rather small. You posted the size of the large tables but it's really the other sizes that are part of the problem
Your problem is from the fact that you can't join the smallest tables first. Your join order should be from small to large. But because your mapping tables don't include a securitygroup-to-university table, you can't join the smallest ones first. So you wind up starting with one or the other, to a big table, to another big table and then with that large intermediate result you have to go to a small table.
If you always have current_univ and current_secgrp and root as inputs you want to use them to filter as soon as possible. The only way to do that is to change your schema some. In fact, you can leave the existing tables in place if you have to but you'll be adding to the space with this suggestion.
You've normalized the data very well. That's great for speed of update... not so great for querying. We denormalize to speed querying (that's the whole reason for datawarehouses (ok that and history)). Build a single mapping table with the following columns.
Univ_id, SecGrp_ID, Root, catalog_id. Make it an index organized table of the first 3 columns as pk.
Now when you query that index with all three PK values, you'll finish that index scan with a complete list of allowable catalog Id, now it's just a single join to the cat table to get the cat item details and you're off an running.
The Oracle cost-based optimizer makes use of all the information that it has to decide what the best access paths are for the data and what the least costly methods are for getting that data. So below are some random points related to your question.
The first three tables that you've listed all have primary keys. Do the other tables (catalog_university and catalog_securitygroup) also have primary keys on them?? A primary key defines a column or set of columns that are non-null and unique and are very important in a relational database.
Oracle generally enforces a primary key by generating a unique index on the given columns. The Oracle optimizer is more likely to make use of a unique index if it available as it is more likely to be more selective.
If possible an index that contains unique values should be defined as unique (CREATE UNIQUE INDEX...) and this will provide the optimizer with more information.
The additional indexes that you have provided are no more selective than the existing indexes. For example, the index on (catalog.id, catalog.name, catalog.owner) is unique but is less useful than the existing primary key index on (catalog.id). If a query is written to select on the catalog.name column, it is possible to do and index skip scan but this starts being costly (and most not even be possible in this case).
Since you are trying to select based in the catalog.root column, it might be worth adding an index on that column. This would mean that it could quickly find the relevant rows from the catalog table. The timing for the second query could be a bit misleading. It might be taking 2 seconds to find 50 matching rows from catalog, but these could easily be the first 50 rows from the catalog table..... finding 50 that match all your conditions might take longer, and not just because you need to join to other tables to get them. I would always use create table as select without restricting on rownum when trying to performance tune. With a complex query I would generally care about how long it take to get all the rows back... and a simple select with rownum can be misleading
Everything about Oracle performance tuning is about providing the optimizer enough information and the right tools (indexes, constraints, etc) to do its job properly. For this reason it's important to get optimizer statistics using something like DBMS_STATS.GATHER_TABLE_STATS(). Indexes should have stats gathered automatically in Oracle 10g or later.
Somehow this grew into quite a long answer about the Oracle optimizer. Hopefully some of it answers your question. Here is a summary of what is said above:
Give the optimizer as much information as possible, e.g if index is unique then declare it as such.
Add indexes on your access paths
Find the correct times for queries without limiting by rowwnum. It will always be quicker to find the first 50 M&Ms in a jar than finding the first 50 red M&Ms
Gather optimizer stats
Add unique/primary keys on all tables where they exist.
The use of rownum is wrong and causes all the rows to be processed. It will process all the rows, assigned them all a row number, and then find those between 0 and 50. When you want to look for in the explain plan is COUNT STOPKEY rather than just count
The query below should be an improvement as it will only get the first 50 rows... but there is still the issue of the joins to look at too:
SELECT ccc.* FROM (
SELECT cc.*, ROWNUM AS n FROM (
SELECT c.id, c.name, c.owner
FROM catalog c
WHERE c.root = 100
ORDER BY name
) cc
where rownum <= 50
) ccc WHERE ccc.n > 0 AND ccc.n <= 50;
Also, assuming this for a web page or something similar, maybe there is a better way to handle this than just running the query again to get the data for the next page.
try to declare a cursor. I dont know oracle, but in SqlServer would look like this:
declare #result
table (
id numeric,
name varchar(255)
);
declare __dyn_select_cursor cursor LOCAL SCROLL DYNAMIC for
--Select
select distinct
c.id, c.name
From [catalog] c
inner join university u
on u.catalog = c.id
and u.university = 300
inner join catalog_securitygroup s
on s.catalog = c.id
and s.securitygroup = 200
Where
c.root = 100
Order by name
--Cursor
declare #id numeric;
declare #name varchar(255);
open __dyn_select_cursor;
fetch relative 1 from __dyn_select_cursor into #id,#name declare #maxrowscount int
set #maxrowscount = 50
while (##fetch_status = 0 and #maxrowscount <> 0)
begin
insert into #result values (#id, #name);
set #maxrowscount = #maxrowscount - 1;
fetch next from __dyn_select_cursor into #id, #name;
end
close __dyn_select_cursor;
deallocate __dyn_select_cursor;
--Select temp, final result
select
id,
name
from #result;